JP4381772B2 - Mold monitoring apparatus, method and program - Google Patents

Description

  The present invention relates to a mold monitoring apparatus, method, and program.

  2. Description of the Related Art Conventionally, in a molding machine such as an injection molding machine, a screw is advanced in a heating cylinder, heated and melted resin is injected at a high pressure, and filled in a cavity of a mold apparatus. A molded product is molded by cooling and solidifying the resin in the cavity. And the technique which acquires the image of the metal mold | die apparatus in the said molding machine with an imaging device, and monitors the metal mold | die during shaping | molding is proposed (for example, refer patent document 1).

In this case, based on the image of the mold apparatus acquired by the imaging apparatus, it can be confirmed that there is a molded product on the mold surface of the movable mold at the time of mold opening, or on the mold surface of the movable mold It is possible to determine the quality of the molded product in the mold and measure the shape, and to confirm that the molded product has been removed from the mold surface of the movable mold when the ejector pin has been ejected. Yes. For example, when it is confirmed that the molded product has been removed from the mold surface of the movable mold when the ejector pin has been ejected, an image of the mold apparatus acquired when the ejector pin has been ejected, that is, a discrimination target image Is obtained in advance and compared with the reference image stored in the storage means, it is determined whether or not there is an abnormality in the determination target image, that is, whether or not there is an image abnormality. When it is determined that there is an image abnormality, it is determined that the molded product has not been removed from the mold surface of the movable mold when the ejector pin has been ejected, that is, it is determined that a molding abnormality has occurred. The operation can be stopped and an alarm can be output.
JP 2002-111772 A

  However, in the conventional mold monitoring apparatus, the luminance of the discrimination target image and the reference image are compared for each pixel, and the number of abnormal pixels whose luminance difference is equal to or greater than a predetermined value is counted. When the number exceeds a threshold value, it is determined that there is an image abnormality. Therefore, a so-called misjudgment that determines that a molding abnormality has occurred even though a molding abnormality has not actually occurred occurs.

  In general, the shape of the cavity is complex, and depending on the location of the cavity, even if an image of the mold apparatus is acquired by the imaging device under the same conditions, the reflected light from the complex surface of the cavity for some reason each time. Changes, and there is a portion where the luminance of the pixel changes. In this case, even if the molding is performed normally, the number of abnormal pixels exceeds the threshold value and it is determined that there is an image abnormality. Therefore, it is determined that a molding abnormality has occurred, and an erroneous determination occurs. If a misclassification occurs, the molding machine operation is stopped or an alarm is output even though the molding is normally performed, so that the throughput of the molding machine decreases.

  However, the part where the luminance of the pixel changes in the mold apparatus, that is, the area causing the erroneous determination is almost constant, and the reproducibility is high. Therefore, based on the experience of the operator, the area is excluded from the mold monitoring range in advance, or abnormal pixels are excluded from the mold monitoring range when it is determined that there is an abnormality in the image to be determined while monitoring the mold being molded. It is conceivable to prevent the occurrence of misjudgment. However, a wealth of experience is required to properly set the areas that cause misclassification.If the areas are set inappropriately, there is an image error even if a molding error actually occurs. And it is not determined that a molding abnormality has occurred. Also, in order to exclude abnormal pixels from the mold monitoring range when it is determined that there is an abnormality in the image to be determined while monitoring the mold during molding, the operation of the molding machine is stopped and the mold monitoring device is set. Needs to be changed, and the throughput of the molding machine is reduced.

  The present invention solves the problems of the conventional mold monitoring apparatus, extracts and registers in advance the area causing misdetermination before starting molding, and the successive normal determination results after starting molding By extracting and registering the area that causes misidentification from the above, it is possible to appropriately prevent the occurrence of misidentification, and it is necessary to stop the molding machine operation and change the setting of the mold monitoring device An object of the present invention is to provide a mold monitoring apparatus, method, and program that are free from defects.

Therefore, in the mold monitoring apparatus of the present invention, an image input unit that acquires an image of a mold surface of a movable mold or a fixed mold of a molding machine, and a reference image acquired in advance by the image input unit are stored. A reference image database, a determination processing unit that compares the determination target image acquired by the image input unit with the reference image, executes a determination process, and determines whether the determination target image is abnormal; which has a misclassified candidate database storing misclassification candidate region extracted by incrementing or decrementing the characteristic values of the observation unit in the determination target image for the characteristic values of the observation unit in the previous determination target image, wherein The discrimination processing unit includes a characteristic value of the observation unit in the discrimination target image excluding the erroneous discrimination candidate area and a characteristic of the observation unit in the reference image. Is compared with a value to determine whether or not each observation unit in the determination target image is an abnormal unit, and the number of the abnormal units or the size of the abnormal unit connected component that combines the abnormal units has a threshold value. If it exceeds, it is determined that the determination target image is abnormal .

  In still another mold monitoring apparatus of the present invention, the determination processing unit corrects the size of the abnormal unit connected component by the misidentification candidate region.

  In still another mold monitoring apparatus of the present invention, the determination processing unit determines whether or not there is an abnormality in the determination target image by excluding the erroneous determination candidate region from the determination target image.

  Still another mold monitoring apparatus of the present invention further includes a post-discrimination processing unit that extracts the erroneous discrimination candidate area based on the result of the discrimination process of the discrimination processing unit.

  In still another mold monitoring apparatus of the present invention, the post-discrimination processing unit further stores the extracted misclassification candidate region in the misclassification candidate database.

  In still another mold monitoring apparatus of the present invention, the post-discrimination processing unit is further configured to perform the discrimination processing unit that is executed in misclassification learning in which an operation that induces an erroneous discrimination before molding of the molding machine is executed. The erroneous discrimination candidate area is extracted based on the result of the discrimination process.

  In still another mold monitoring apparatus of the present invention, the post-discrimination processing unit further determines the erroneous discrimination candidate area based on a discrimination process result of the discrimination processing unit executed during molding of the molding machine. Extract.

In the mold monitoring method of the present invention, the image input unit acquires an image of the mold surface of the movable mold or the fixed mold of the molding machine, and the reference image acquired in advance by the image input unit is stored in the reference image database. The discrimination processing unit compares the discrimination target image acquired by the image input unit with a reference image acquired in advance, and executes discrimination processing to determine whether the discrimination target image is abnormal. , discrimination post-processing unit extracts misclassification candidate region by incrementing or decrementing the characteristic values of the observation unit in the determination target image for the characteristic values of the observation unit in the previous determination target image, the misclassified candidate stores the misclassification candidate database area, the determination processing section has a characteristic value of the observation unit in the determination target image excluding the erroneous determination candidate region and the The characteristic value of the observation unit in the quasi-image is compared to determine whether or not each observation unit in the discrimination target image is an abnormal unit, and the number of the abnormal units or the abnormal units obtained by combining the abnormal units When the size of the connected component exceeds the threshold, it is determined that the determination target image is abnormal .

  In another mold monitoring method of the present invention, when obtaining the discrimination target image, the discrimination target image is obtained by changing the surrounding environment while the mold is stationary.

  In still another mold monitoring method of the present invention, when the discrimination target image is acquired, the discrimination target image is acquired by changing the position of the movable part of the mold while the mold is stationary.

In the mold monitoring program of the present invention, a computer for mold monitoring, an image input unit for acquiring an image of a movable mold or a fixed mold of a molding machine, and a reference image acquired in advance by the image input unit And a discrimination processing unit that compares the discrimination target image acquired by the image input unit with the reference image to execute discrimination processing and discriminates whether or not the discrimination target image is abnormal When, together to function as a misclassification candidate database storing misclassification candidate region extracted by incrementing or decrementing the characteristic values of the observation unit in the determination target image for the characteristic values of the observation unit in the previous determination target image The discrimination processing unit includes a characteristic value of an observation unit in the discrimination target image excluding the misclassification candidate region and the reference Comparing the characteristic value of the observation unit in the image to determine whether or not each observation unit in the image to be discriminated is an abnormal unit, and the number of the abnormal units or the abnormal unit concatenation of the abnormal units When the component size exceeds the threshold, it is determined that the determination target image is abnormal .

According to the present invention, a mold monitoring apparatus stores an image input unit that acquires an image of a mold surface of a movable mold or a fixed mold of a molding machine, and a reference image acquired in advance by the image input unit. A reference image database, a determination processing unit that compares the determination target image acquired by the image input unit with the reference image, executes a determination process, and determines whether or not the determination target image is abnormal; and A misclassification candidate database storing a misclassification candidate region extracted by incrementing or decrementing the characteristic value of the observation unit in the discrimination target image with respect to the characteristic value of the observation unit in the previous discrimination target image , and the discrimination The processing unit compares the characteristic value of the observation unit in the determination target image excluding the erroneous determination candidate region with the characteristic value of the observation unit in the reference image. Determining whether each observation unit in the determination target image is an abnormal unit, the number of the abnormal units, or the size of the abnormal unit connected component that combines the abnormal units exceeds the threshold It is determined that there is an abnormality in the determination target image .

  In this case, it is possible to extract and register a region that causes a misclassification in advance before starting molding, and sequentially extract and register a region that causes a misclassification even after molding starts. Therefore, it is possible to appropriately prevent the occurrence of erroneous determination, and it is not necessary to change the setting of the mold monitoring device by stopping the operation of the molding machine.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. The mold monitoring apparatus of the present invention can be applied to any type of molding machine, but here, a case where it is applied to an injection molding machine will be described.

  FIG. 2 is a schematic diagram of the molding machine according to the first embodiment of the present invention, and FIG. 3 is a block diagram showing the equipment configuration of a mold monitoring apparatus applied to the molding machine according to the first embodiment of the present invention. .

  In FIG. 2, 10 is an injection molding machine as a molding machine. Reference numeral 17 denotes a frame, 11 a fixed platen as a fixed mold support device fixed to the frame 17, and 15 a predetermined distance between the fixed platen 11 and movable relative to the frame 17. It is a toggle support as a base plate disposed in the box. The toggle support 15 functions as a toggle type mold clamping device support device. Reference numeral 16 denotes a plurality of, for example, four, tie bars provided between the fixed platen 11 and the toggle support 15 as guide means.

  Reference numeral 14 denotes a movable platen as a movable mold support device which is disposed so as to face the fixed platen 11 and which can be moved back and forth (moved in the left-right direction in FIG. 2) along the tie bar 16. The injection molding machine 10 has a mold apparatus including a fixed mold 12 and a movable mold 13. A fixed mold 12 is attached to a mold attachment surface of the fixed platen 11 facing the movable platen 14, and a movable mold 13 is attached to a mold attachment surface of the movable platen 14 facing the fixed platen 11. Is attached.

  A drive device for moving an ejector pin of an ejector device (not shown) may be attached to the rear end (left end in FIG. 2) of the movable platen 14.

  A toggle mechanism 20 as a toggle type clamping device of the injection molding machine 10 is attached between the movable platen 14 and the toggle support 15, and a rear end (left end in FIG. 2) of the toggle support 15 is attached. A mold clamping motor 26 is provided as a mold clamping drive source for operating the toggle mechanism 20. The mold clamping motor 26 includes a motion direction conversion device (not shown) composed of a ball screw mechanism or the like that converts rotational motion into reciprocating motion, and moves the drive shaft 25 forward and backward (moves in the left-right direction in FIG. 2). The toggle mechanism 20 can be activated.

  Here, the toggle mechanism 20 is swingably attached to the crosshead 24 attached to the drive shaft 25, the second toggle lever 23 attached to the crosshead 24 so as to be able to swing, and the toggle support 15. The first toggle lever 21 and the toggle arm 22 that is swingably attached to the movable platen 14. Then, the first toggle lever 21 and the second toggle lever 23 and the first toggle lever 21 and the toggle arm 22 are linked with each other. The toggle mechanism 20 is a so-called inner winding five-node double toggle mechanism, and has a vertically symmetrical configuration.

  The toggle mechanism 20 can be operated by driving the mold clamping motor 26 and moving the cross head 24 as a driven member forward and backward (moving in the left-right direction in FIG. 2). As a result, the movable mold 13 moves relative to the fixed mold 12, and mold closing, mold clamping, and mold opening are performed. That is, when the cross head 24 is moved forward (moved in the right direction in FIG. 2), the movable platen 14 and the movable mold 13 are moved forward to perform mold closing and clamping. Further, when the cross head 24 is moved backward (moved leftward in FIG. 2), the movable platen 14 and the movable mold 13 are moved backward, and the mold is opened. A mold clamping force obtained by multiplying the propulsive force of the mold clamping motor 26 by a toggle magnification is generated, and the mold clamping is performed by the mold clamping force.

  The toggle support 15 is formed with a plurality of, for example, four tie bar insertion holes (not shown), and the left end of the tie bar 16 in FIG. 2 is inserted into each of the tie bar insertion holes. A right end (not shown) of the tie bar 16 is fixed to the fixed platen 11 by a fixing nut 16a. The tie bar 16 has a screw portion 16b having a screw formed on the outer periphery of the left end in FIG. 2, and an adjustment nut 27 is screwed onto the screw portion 16b of each tie bar 16. The adjustment nut 27 is attached to the rear end of the toggle support 15 so as to be rotatable and immovable in the axial direction of the tie bar 16.

  3, 30 is an injection device of the injection molding machine 10, and has a heating cylinder 31 provided with an injection nozzle 32 at the tip (left end in FIG. 3). Further, a raw material hopper 34 for introducing raw resin such as resin pellets into the heating cylinder 31 is attached to the heating cylinder 31. Further, a screw is disposed in the heating cylinder 31 so as to be able to rotate and to advance and retreat. Then, behind the heating cylinder 31 (right side in FIG. 3), an injection driving device for rotating the screw in the heating cylinder 31 and moving forward or backward (moving leftward or rightward in FIG. 3). 33 is arranged.

  In the metering step, the injection driving device 33 is driven to rotate the screw and store (accumulate) resin in front of the screw, so that the screw is retracted to a predetermined position. At this time, the raw material resin supplied from the raw material hopper 34 is heated in the heating cylinder 31 and melted, and the screw is retracted by the pressure of the resin stored in front of the screw (left side in FIG. 3). .

  Subsequently, in the injection process, the injection nozzle 32 at the tip of the heating cylinder 31 is inserted into the nozzle insertion hole 11 a of the fixed platen 11 and pressed against the fixed mold 12. Then, the injection driving device 33 is driven to advance the screw. Note that the mold is closed before the injection process, and the mold surface of the movable mold 13 is in contact with the mold surface of the fixed mold 12. Therefore, a cavity space having a shape corresponding to the shape of the molded product is formed by the cavity 12 a of the fixed mold 12 and the cavity 13 a of the movable mold 13. In this state, as the screw moves forward, the resin accumulated in front of the screw is injected from the injection nozzle 32 and filled into the cavity space through the sprue 12b formed in the fixed mold 12. Is done.

  After the mold is clamped, the resin in the cavity space is cooled and solidified to some extent, and when the molded product is molded, the movable platen 14 and the movable mold 13 are retracted and the mold is opened. . The movable platen 14 includes an ejector device for projecting a molded product from the mold surface of the movable mold 13, that is, ejecting the molded product. In the one-shot molding process, when the mold apparatus is opened, the molded product adheres to the cavity 13a on the mold surface of the movable mold 13, so that the ejector apparatus is operated to eject the ejector pin. Is protruded from the surface of the cavity 13a, and the molded product adhering to the surface is ejected and released. The cavities 12a and cavities 13a may be singular or plural. When there are a plurality of molded products, a plurality of molded products can be molded by a one-shot molding process.

  In the present embodiment, the imaging device 41 is attached to the upper part of the fixed platen 11 via a mounting bracket 41a, and acquires an image of a mold surface including the cavity 13a of the movable mold 13. The imaging device 41 may be attached to any member as long as it is disposed at a position where an image of the mold surface of the movable mold 13 can be obtained from above the fixed mold 12. It may be attached to the upper part of the fixed mold 12. The image pickup apparatus 41 preferably has a case body that is waterproofed. The case includes a lens, a charge coupled device (CCD), and a complementary metal oxide (CMOS). An imaging device such as a semiconductor) or an imaging tube is provided. The imaging device 41 may be of any kind as long as it can acquire a still image or a moving image, and may be an industrial camera, a surveillance camera, or the like. The imaging device 41 is arranged so that a range including at least the cavity 13a of the mold surface of the movable mold 13 is an imaging range, that is, a visual field range, and the movable mold in a state where the mold is opened. Images of 13 mold surfaces are acquired.

  In addition, an illumination unit 42 is disposed in the vicinity of the imaging device 41 as shown in FIG. The illumination unit 42 includes a lamp such as a halogen lamp attached to a lamp holder, and is fixed to an upper portion of the fixed platen 11 and an upper portion of the fixed mold 12 through a fixing bracket (not shown). The mold surface of the movable mold 13 is irradiated from above.

  The mold monitoring apparatus in the present embodiment has a control device 40. The control device 40 includes a control substrate 43, an I / O substrate 44, an I / O conversion substrate 45, an image processing substrate 46, a display switching unit 47, and a display device 48. Here, the control substrate 43 is a control device for controlling the overall operation of the mold monitoring device such as a calculation means such as a CPU or MPU, a storage means such as a semiconductor memory or a magnetic disk, and the operation of the injection molding machine 10. Is a board made of a mother board, a daughter board or the like on which is mounted.

  The I / O board 44 includes an input / output interface, a network card, a communication device, and the like, and is connected to an input device such as a keyboard, a mouse, a joystick, a touch panel, and the like. A signal input process and an output signal output process from the control board 43, that is, an in-out (I / O) process of the control board 43 is executed.

  Further, the I / O conversion board 45 is a board on which arithmetic means, storage means, and the like are mounted, and converts an output signal output from the control board 43 via the I / O board 44 into an image signal. To the image processing substrate 46. The I / O conversion board 45 converts the image signal output from the image processing board 46 into a normal input signal and inputs it to the control board 43 via the I / O board 44. That is, the I / O conversion board 45 performs input / output signal conversion processing between the control board 43 and the image processing board 46.

  The image processing board 46 is a board on which image processing arithmetic means, image processing storage means such as a VRAM, and the like are mounted. Image processing of the image signal received from the imaging device 41, and Image processing of the output signal from the control board 43 received via the I / O board 44 and the I / O conversion board 45 is executed. The display switching unit 47 switches the output signal from the control board 43 and the image signal from the image processing board 46 received via the I / O board 44 and outputs them to the display device 48. Further, the display device 48 includes a CRT, a liquid crystal display, an LED (Light Emitting Diode) display, and the like. The display device 48 receives an output signal from the control board 43 and an image from the image processing board 46 received via the display switching unit 47. Display the signal. The display device 48 is preferably capable of color display.

  As described above, the control device 40 is a kind of computer, and performs image processing of the image acquired by the imaging device 41, that is, the discrimination target image, and compares it with the reference image stored and registered in advance in the storage unit. Thus, it can be determined whether or not the determination target image is normal. In addition, the said control apparatus 40 may be comprised independently, and may be comprised integrally with another control apparatus etc. For example, it may be one of a plurality of systems constructed in the control device of the injection molding machine 10.

  In the present embodiment, the imaging device 41 obtains a signal indicating that the mold device has been opened from the control device of the injection molding machine 10 and the movable mold at the time when the mold device is opened. The mold surface of the mold 13 is imaged. When the mold apparatus is opened, the molded product is attached to the cavity 13a on the mold surface of the movable mold 13, and includes at least the cavity 13a on the mold surface of the movable mold 13. Is in the field of view of the imaging device 41, the molded image of the molded product is acquired by the imaging device 41.

  Here, the control device 40 performs image processing using the image of the molded product acquired by the imaging device 41 as a discrimination target image, and compares the image with a reference image of the molded product stored and registered in advance in the storage unit. When determining whether or not a molded product is good, it is possible to determine whether or not burrs or shorts have occurred in the molded product. Then, in the molding process, if the molten resin leaks and burrs are generated, or if the molten resin is insufficient and a short circuit occurs, it is determined that the product is defective, and if not, it is determined to be a good product. For example, the quality of the molded product may be determined by determining the degree of warpage of the molded product by digitizing the inclination of the sprue part attached to the molded product by image processing. In this case, if the inclination of the sprue portion is equal to or greater than a predetermined value, the degree of warping is large, so that it is determined to be a defective product, and otherwise, it is determined to be a good product.

  When it is determined that the product is defective, the control device 40 displays on the display device 48 that the product is defective. In the case where the control device 40 includes sound output means such as a speaker, a warning sound may be generated. When the injection molding machine 10 has a molded product take-out device, the determination result that the control device 40 is a defective product is transmitted to the control device of the injection molding machine 10, and the molded product take-out device is sent to the control device. The molded product determined to be defective can be removed by operating the.

  Further, the control device 40 can perform the residual determination of the molded product for confirming that the molded product does not remain in the fixed mold 12 or the movable mold 13. In this case, the control device 40 performs image processing using the image of the mold surface in the movable mold 13 acquired by the imaging device 41 at the time when the mold apparatus is opened, as a discrimination target image, and moves the movable mold. It is determined whether or not the molded product is attached to all the cavities 13a. When no molded product is attached to any of the cavities 13a, it is determined that the molded product remains in the fixed mold 12. Further, the control device 40 performs image processing using the image of the mold surface in the movable mold 13 acquired by the imaging device 41 immediately after the ejector device is operated as a discrimination target image, and from all the cavities 13a. It is determined whether or not the molded product is ejected. When no molded product is ejected from any of the cavities 13a, it is determined that the molded product remains in the movable mold 13.

  When the molded product remains in the fixed mold 12 or the movable mold 13, when the next shot is molded, the molded product is not only damaged or defective, but also the mold apparatus There is a risk of damage. Therefore, when it is determined that the molded product remains in the fixed mold 12 or the movable mold 13, the control device 40 displays on the display device 48 that the molded product remains. In the case where the control device 40 includes sound output means such as a speaker, a warning sound may be generated. In addition, the determination result that the molded product of the control device 40 remains may be transmitted to the control device of the injection molding machine 10 so that the control device stops the operation of the injection molding machine 10.

  Next, the configuration of the control device 40 in the present embodiment will be described in detail. Here, description will be made assuming that the control device 40 also functions as a control device of the injection molding machine 10.

  FIG. 1 is a block diagram showing the configuration of the control device of the mold monitoring apparatus according to the first embodiment of the present invention.

  In the present embodiment, the control device 40 includes a storage unit 50 that stores various data, a calculation unit 60 that executes various processes, and a control device for the injection molding machine 10, as shown in FIG. A molding machine control unit 63 that functions as: Further, the control device 40 includes a display unit 66 for displaying various information such as images and characters, an image input unit 67 for inputting an image such as an image of a mold surface in the movable mold 13, and a setting necessary for determination. A setting input unit 68 for inputting various setting information such as values and an operation instruction unit 69 for instructing the operation of the injection molding machine 10 are connected. The display device 48 corresponds to the display unit 66, the imaging device 41 corresponds to the image input unit 67, and input devices such as a keyboard, a mouse, a joystick, and a touch panel connected to the I / O board 44 are setting input units. 68 and the operation instruction unit 69.

  The storage unit 50 includes an input image database 51 that stores an image input by the image input unit 67 and a setting information database 54 that stores setting information input by the setting input unit 68. Further, the input image database 51 includes a reference image database 52 that stores a reference image acquired in advance by the imaging device 41, and a captured image database 53 that stores images such as a discrimination target image acquired by the imaging device 41. Prepare. In addition, the setting information database 54 stores a misclassification candidate database 55 that stores a misclassification candidate region that causes a misclassification in a discrimination process, a sensitivity database 56 that stores setting information regarding sensitivity, and setting information regarding a monitoring range. A monitoring range database 57 and a discrimination size database 58 for storing setting information regarding the discrimination size are provided.

  And the said calculating part 60 is provided with the discrimination | determination process part 61 which performs a discrimination | determination process, and the post-determination process part 62 which performs processes after discrimination | determination, such as misclassification learning. In this case, the determination processing unit 61 performs a determination process by comparing the determination target image acquired by the imaging device 41 with a reference image, determines whether or not the determination target image is abnormal, The post-discrimination processing unit 62 extracts a misclassification candidate region based on the result of the discrimination process of the discrimination processing unit 61 and stores the misclassification candidate region in the misclassification candidate database 55. Further, the molding machine control unit 63 receives a timing signal indicating that the mold opening operation has been completed, the ejector pin projecting operation has been completed, or the like from the sensors disposed in each part of the injection molding machine 10. Timing signals indicating that the mold closing operation can be started and the ejector pin can be protruded are output to the signal input unit 64 and the driving source such as the motor of each part of the injection molding machine 10. A timing signal output unit 65 is provided.

  Next, the operation of the mold monitoring apparatus having the above configuration will be described. First, a basic operation for determining whether there is an abnormality in the determination target image will be described.

  FIG. 4 is a diagram showing a mold surface of the movable mold according to the first embodiment of the present invention and showing a state where a molded product is attached, and FIG. 5 is a movable mold according to the first embodiment of the present invention. FIG. 6 is a diagram showing a state where the molded product is released from the mold surface, FIG. 6 is a diagram showing the luminance of the area A in FIG. 4 in the first embodiment of the present invention, and FIG. FIG. 8 is a diagram showing a method for calculating the size of an abnormal unit connected component of a discrimination target image of a mold surface of a movable mold in the first embodiment, and FIG. 8 shows the movable mold in the first embodiment of the present invention. It is a flowchart which shows the operation | movement which discriminate | determines abnormality of the discrimination | determination target image of a mold surface.

  When a molded product is molded in the one-shot molding process of the injection molding machine 10, the movable platen 14 and the movable mold 13 are retracted, and the mold is opened. When the movable mold 13 stops at the most retracted position, a timing signal indicating that the mold opening operation is completed is input to the timing signal input unit 64 from the sensor of the injection molding machine 10. Then, the imaging device 41 acquires an image of the mold surface including the cavity 13a of the movable mold 13 as shown in FIG. In this case, a molded product 71 is attached to the cavity 13a of the movable mold 13 as shown in FIG. In FIG. 4, 13b indicates a hole into which the guide pin is inserted. The acquired image is stored in the captured image database 53 as a discrimination target image.

  Subsequently, the ejector device is operated to cause the ejector pin to protrude from the surface of the cavity 13a, and the molded product 71 adhering to the surface is ejected and released. Then, a timing signal indicating that the ejecting operation of the ejector pin is completed from the sensor of the injection molding machine 10 is input to the timing signal input unit 64. Then, the imaging device 41 acquires an image of the mold surface including the cavity 13a of the movable mold 13 as shown in FIG. In FIG. 5, reference numeral 13c denotes a hole from which the ejector pin protrudes. In this case, nothing is attached to the cavity 13a of the movable mold 13 as shown in FIG. Also in this case, the acquired image is stored in the captured image database 53 as a discrimination target image.

  Then, the control device 40 obtains the discrimination target image at the time when the mold opening operation is completed and / or the discrimination target image at the time when the ejector pin protrusion operation is completed by the imaging device 41 in advance and acquires the reference image. Each of the images is compared with the reference image stored in the database 52 to determine whether there is an abnormality in the determination target image, that is, whether there is an image abnormality.

  Here, the image acquired by the imaging device 41 is a digital image that is digitized and configured of a plurality of pixels. For example, in the case of a horizontally long image, the image is divided into 640 pixels in the horizontal direction and 480 pixels in the vertical direction, and is composed of 307200 pixels. That is, the number of pixels is 307200. Then, the determination processing unit 61 sets a predetermined range in the determination target image as the monitoring range, and compares the characteristic value of the pixel included in the monitoring range with the characteristic value of the pixel included in the monitoring range of the reference image. The monitoring range is, for example, the range shown in FIG. Further, the characteristic value of the pixel may be of any kind, but here, a case where luminance indicating brightness is used as the characteristic value will be described.

  In this case, the pixels included in the A portion within the monitoring range in the discrimination target image have luminance as shown in FIG. 6, for example. In FIG. 6, the vertical axis (y direction) is assigned 176 to 185 on the vertical axis, and the horizontal (x direction) addresses are assigned 248 to 258 on the horizontal axis. Each frame corresponds to a pixel, and the number in each frame indicates the luminance of each pixel. The luminance is displayed in 256 levels, with 0 for black and 255 for white.

  The discrimination processing unit 61 uses each pixel as an observation unit, and when the difference between the luminance of the observation unit in the discrimination target image and the luminance of the corresponding observation unit in the reference image is equal to or greater than the abnormal unit discrimination threshold, the discrimination processing unit 61 The observation unit in the target image is determined as an abnormal unit. Here, the abnormal unit discrimination threshold is set in advance. As will be described later, a plurality of adjacent pixels are grouped to form an observation unit composed of a plurality of pixels, and the luminance of the observation unit in the discrimination target image is different from the luminance of the corresponding observation unit in the reference image. May be determined as an abnormal unit in the determination target image.

  Subsequently, the determination processing unit 61 determines whether an observation unit around the abnormal unit is also an abnormal unit, and if it is an abnormal unit, combines them. That is, when the hatched observation unit shown in FIG. 7 (a) is an abnormal unit, it is determined whether or not the observation units assigned with the numbers 1 to 8 in the vicinity thereof are also abnormal units, In the case of an abnormal unit, an abnormal unit connected component is created by combining the observation units that are hatched. Subsequently, when the size of the abnormal unit connected component exceeds the component size determination threshold, the determination processing unit 61 determines that there is an abnormality in the determination target image, that is, there is an image abnormality. The component size determination threshold value is input in advance from the setting input unit 68 by the operator and stored in the determination size database 58. For example, in the example shown in FIG. 7B, when the hatched observation unit is an abnormal unit, the size of the abnormal unit connected component is 2, 1, 1, and 16 from the upper left to the lower right. Therefore, if the component size determination threshold is less than 16, the example shown in FIG. 7B is determined to have an image abnormality.

  The determination processing unit 61 may determine whether or not there is an abnormality in the determination target image based on the number of abnormal units without creating the abnormal unit connected component. In this case, when the number of abnormal units exceeds a preset abnormal unit threshold, the determination processing unit 61 determines that there is an abnormality in the determination target image, that is, there is an image abnormality. The abnormal unit threshold value is previously input by the operator from the setting input unit 68 and stored in the discrimination size database 58.

Next, a flowchart will be described.
Step S1 The timing signal is input to the timing signal input unit 64.
Step S2 The imaging device 41 captures an image of the mold surface including the cavity 13a of the movable mold 13.
Step S3 The discrimination processing unit 61 configures each pixel as an observation unit.
Step S4: A determination is made for each observation unit.
Step S5: Calculate the number of abnormal units or create an abnormal unit connected component size.
Step S6: It is determined whether or not the number of abnormal units exceeds an abnormal unit threshold, or whether or not an abnormal unit connected component size determination threshold is exceeded. If the abnormal unit number exceeds the abnormal unit threshold value, or if the abnormal unit connected component size determination threshold value is exceeded, the process proceeds to step S7. If the abnormal unit number does not exceed the abnormal unit threshold value, or the abnormal unit connected component If the size determination threshold is not exceeded, the process proceeds to step S8.
Step S7: It is determined that there is an image abnormality, and the process is terminated.
Step S8: It is determined that there is no image abnormality, and the process is terminated.

  Next, an operation of misclassification learning in which a misclassification candidate region is extracted and registered in advance will be described.

  FIG. 9 is a diagram showing a method for configuring the observation unit by grouping the pixels of the discrimination target image on the mold surface of the movable mold in the first embodiment of the present invention, and FIG. 10 is the first embodiment of the present invention. It is a figure which shows the change of the buffer of the discrimination target image of the mold surface of the movable mold in the embodiment.

  In the present embodiment, post-determination processing is performed to prevent the occurrence of so-called misjudgment, in which it is determined that a molding abnormality has occurred even though no molding abnormality has actually occurred. The unit 62 extracts a region that is likely to cause a misclassification in the discrimination process executed by the discrimination processing unit 61 as a misclassification candidate region and stores it in the misclassification candidate database 55. Usually, the region that is likely to cause misidentification is a region in which light around the injection molding machine 10 is regularly reflected by the mold surface of the movable mold 13 and enters the imaging device 41, and changes in time or periodically. It is the outline | summary and internal area | region of movable parts, such as a guide pin and an ejector pin, of a mold apparatus to perform.

  Therefore, in the present embodiment, misclassification learning is performed before the injection molding machine 10 performs actual molding, that is, before actual mold monitoring. In the misclassification learning, the discrimination processing unit 61 discriminates whether or not there is an image abnormality while performing an operation that causes misclassification, and the post-discrimination processing unit 62 causes the image abnormality based on the discrimination result. An area is extracted and stored in the misclassification candidate database 55 as a misclassification candidate area.

  Here, when the discrimination processing unit 61 and the post-discrimination processing unit 62 compare the discrimination target image with the reference image, a description will be given of a case where a plurality of adjacent pixels are combined to form an observation unit composed of a plurality of pixels. To do. For example, as shown in FIG. 9A, the adjacent 2 × 2, that is, four pixels of the discrimination target image divided into 640 pixels in the horizontal direction and 480 pixels in the vertical direction are combined into one observation unit. By configuring and re-dividing, it is possible to obtain a discrimination target image after re-division as shown in FIG. Also, the luminance value of the observation unit after the average values of the luminance values of the four adjacent pixels are collected. 9A shows the state of the pixel of the digital image acquired by the imaging device 41, and FIG. 9B shows a state after re-division into which the horizontal is divided into 320 observation units and the vertical is divided into 240 observation units. A discrimination target image is shown. FIG. 9A is drawn with some pixels omitted, and FIG. 9B is drawn with some observation units omitted.

  In this case, the reference image is also divided into 320 observation units in the horizontal direction and 240 observation units in the vertical direction, as in the discrimination target image after re-division as shown in FIG. 9B. The same applies to the erroneous determination candidate areas stored in the erroneous determination candidate database 55. In this way, the number of observation units of the discrimination target image and the reference image can be reduced by combining a plurality of adjacent pixels to form an observation unit composed of a plurality of pixels. It is possible to reduce the processing load for determining whether there is an image abnormality. Therefore, the processing time for the determination can be shortened. Note that it is desirable not to perform re-division when high-level discrimination based on high resolution is required.

  First, before performing actual mold monitoring, the movable mold 13 is stopped at the most retracted position, and misclassification learning is performed in a state where the movable mold 13 is stationary. In this case, the imaging device 41 repeatedly acquires the image of the mold surface of the movable mold 13 a plurality of times while repeatedly performing an operation that causes erroneous determination, and uses each of the acquired plurality of images as a determination target image. The discrimination processing unit 61 discriminates an abnormal unit. Here, the operation that causes the erroneous determination is, for example, an operation in which ambient light is changed by repeatedly turning on and off the ceiling illumination in a place where the injection molding machine 10 is disposed. In addition, you may repeat and perform the operation | movement which shields a metal mold apparatus from ambient light using a light shielding member.

  Then, the post-determination processing unit 62 sequentially adds and accumulates the abnormal units determined by the determination processing unit 61 in the buffer memory, and extracts misdetermination candidate regions based on the accumulation results after a predetermined number of determinations. Store in the misclassification candidate database 55. To describe in more detail with reference to FIG. 10, for example, it is assumed that FIG. 10A-1 shows the result of the determination processing unit 61 determining the abnormal unit based on the first determination target image. Here, the observation unit to which the numerical value 1 is given, that is, the observation units with the addresses 2-2, 3-2 and 4-2 indicated by xy are abnormal units. Then, the post-determination processing unit 62 adds the abnormal unit determined by the determination processing unit 61 to the buffer memory. It is assumed that the observation unit whose numerical value is not shown has a numerical value of 0.

  Here, as shown in FIG. 10B-1, the buffer memory includes the same number of observation units as the discrimination target image. In the initial state of the buffer memory, no data is stored, that is, all observation units do not contain numbers. It is assumed that the observation unit whose numerical value is not shown has a numerical value of 0. Then, when the abnormal unit determined by the determination processing unit 61 as shown in FIG. 10 (a-1) is added to the buffer memory, the buffer memory becomes as shown in FIG. 10 (b-1). . In this case, since no numbers are entered in all observation units in the buffer memory in the initial state, the buffer memory shown in FIG. 10 (b-1) to which the abnormal units are added is shown in FIG. 10 (a-1). It is a copy of the results shown.

  Next, assume that the result of the discrimination processing unit 61 discriminating abnormal units based on the second discrimination target image is shown in FIG. In this case, only the observation unit whose address is 2-2 is an abnormal unit. Then, after the discrimination processing unit 61 adds the result of the discrimination of the abnormal unit to the buffer memory, the post-discrimination processing unit 62 adds the address in the buffer memory to 2-2 as shown in FIG. The observation unit value is incremented (increased) by one. In this case, the values of other observation units in the buffer memory are not changed.

  Next, assume that the result of the discrimination processing unit 61 discriminating abnormal units based on the third discrimination target image is shown in FIG. In this case, only the observation unit whose address is 4-4 is an abnormal unit. Then, after the discrimination processing unit 61 adds the result of discrimination of the abnormal unit to the buffer memory, the post-discrimination processing unit 62 adds the address in the buffer memory to 4-4 as shown in FIG. 10 (b-3). The observation unit value is incremented by one. In this case, the values of other observation units in the buffer memory are not changed.

  Thereafter, the same operation is repeated n times, and when n reaches a predetermined number, that is, when a predetermined number of determination results are stored in the buffer memory, the determination post-processing unit 62 stores the determination results in the buffer memory. , And a misclassification candidate region is extracted based on the accumulation result and stored in the misclassification candidate database 55. Here, the n can be arbitrarily set by inputting from the setting input unit 68, and may be any number, for example, 1. In addition, the post-determination processing unit 62 extracts an observation unit whose numerical value accumulated in the buffer memory is equal to or greater than a misdetermination candidate threshold as an erroneous determination candidate region. Here, the threshold value for misidentification candidates can be arbitrarily set by inputting from the setting input unit 68, and may be any number. Further, the operator can appropriately input a command for ending accumulation of the determination result in the buffer memory from the setting input unit 68.

  For example, in the case where a misclassification candidate area is extracted based on the accumulation result of the buffer memory as shown in FIG. 10B-3, if the misclassification candidate threshold is 1, the address is 2-2, 3 -2, 4-2, and 4-4 observation units are extracted as misclassification candidate areas, and if the misclassification candidate threshold is 2, only the observation unit whose address is 2-2 is extracted as a misclassification candidate area. Is done. Note that the misclassification candidate threshold is input and set in advance from the setting input unit 68 before performing misclassification learning.

  In addition, when extracting the misclassification candidate area, several different values are input from the setting input unit 68 as the misclassification candidate threshold value, and the operator grasps the misclassification candidate area extracted corresponding to each value. Later, an appropriate value can be finally set as a misclassification candidate threshold. In this case, it is desirable that the misidentification candidate area extracted corresponding to each value is displayed on the display unit 66. It is also possible to display such that the color of the observation unit changes according to the numerical value stored in the buffer memory. Thereby, the operator can easily grasp the misidentification candidate regions extracted corresponding to the threshold values for misidentification candidates having several different values. Therefore, based on experience, the operator selects a threshold value for the misclassification candidate so that only observation units corresponding to areas that are considered to be appropriate as areas that cause misclassification are extracted as misclassification candidate areas. Can be set.

  Thereby, the misclassification learning is completed. Then, after misclassification learning is performed and a misclassification candidate region is extracted and stored and registered in the misclassification candidate database 55, a normal molding process of the injection molding machine 10 is started, and the mold monitoring device The operation as described above is performed to determine whether or not there is an abnormality in the determination target image acquired for each molding process, that is, whether or not there is an image abnormality. In this case, the discrimination processing unit 61 discriminates the presence / absence of an image abnormality with respect to the discrimination target image excluding the misdiscrimination candidate area registered in the misdiscrimination candidate database 55.

  As described above, in this embodiment, before starting the normal molding process of the injection molding machine 10, misclassification learning is performed and misclassification candidate areas are registered. In the mold monitoring in the normal molding process of the injection molding machine 10, the presence / absence of an image abnormality is discriminated for the discrimination target image excluding the registered erroneous discrimination candidate area. Therefore, occurrence of erroneous determination can be prevented appropriately, and the injection molding machine 10 is not unnecessarily stopped. Moreover, it is not necessary to stop the operation of the injection molding machine 10 and change the setting of the mold monitoring device.

  Next, a second embodiment of the present invention will be described. In addition, about what has the same structure as 1st Embodiment, the description is abbreviate | omitted by providing the same code | symbol. The description of the same operation and the same effect as those of the first embodiment is also omitted.

  As regions that are likely to cause erroneous determination in the determination processing executed by the determination processing unit 61, there are contours of movable parts such as guide pins and ejector pins of the mold apparatus that change in time or periodically, and regions inside the regions. Therefore, in the present embodiment, as an operation that induces misclassification performed by misclassification learning before the injection molding machine 10 performs actual molding, an operation of slightly projecting or retracting the guide pin or the ejector pin is performed. It is like that. Then, while repeatedly performing such an operation that causes erroneous determination, the imaging device 41 repeatedly acquires the image of the mold surface of the movable mold 13 a plurality of times, and determines each of the acquired plurality of images as a determination target. As an image, the discrimination processing unit 61 discriminates an abnormal unit. Thereby, the observation unit centering on the outline of the guide pin or the ejector pin is determined as an abnormal unit. The operation of sequentially adding and accumulating the abnormal units determined by the determination processing unit 61 to the buffer memory is the same as in the first embodiment, and thus the description thereof is omitted.

  When the discrimination results are stored in the buffer memory, the discrimination post-processing unit 62 sets the value of the observation unit other than 0 as the value stored in the buffer memory to 1, and sets the value of the observation unit as 0 when the stored value is 0. A binary image is created with a value of 0. Then, the post-discrimination processing unit 62 applies known Snakes as a contour extraction method to the binary image, and recognizes an observation unit corresponding to the contour of the guide pin or ejector pin. Subsequently, the post-discrimination processing unit 62 extracts the observation unit corresponding to the contour and the observation unit corresponding to the inside of the contour as a misclassification candidate region, and stores and registers it in the misclassification candidate database 55.

  As described above, in the present embodiment, the contour of the movable part such as the guide pin and the ejector pin of the mold apparatus, which is likely to cause misidentification, and the inner area are appropriately extracted and registered as misidentification candidate areas. Can do.

  Next, a third embodiment of the present invention will be described. In addition, about what has the same structure as 1st and 2nd embodiment, the description is abbreviate | omitted by providing the same code | symbol. Also, the description of the same operations and effects as those of the first and second embodiments is omitted.

  FIG. 11 is a diagram showing a change in the buffer of the discrimination target image on the mold surface of the movable mold in the third embodiment of the present invention.

  In the present embodiment, misclassification learning is performed while the injection molding machine 10 is performing actual molding, that is, while performing actual mold monitoring. In this case, in the mold monitoring in the normal molding process of the injection molding machine 10, the post-determination processing unit 62 sequentially adds and accumulates the abnormal units determined by the determination processing unit 61 in the buffer memory, and the predetermined number of times. Based on the accumulation result after the discrimination, a misclassification candidate region is extracted and stored in the misclassification candidate database 55. To explain in more detail with reference to FIG. 11, for example, it is assumed that the result of the discrimination processing unit 61 discriminating the abnormal unit based on the first discrimination target image is FIG. 11A-1. In the present embodiment, the discrimination processing unit 61 assigns a numerical value 5 to the observation unit that is discriminated to be an abnormal unit. Moreover, the discrimination | determination processing part 61 assign | provides the numerical value 2 to the observation unit discriminate | determined to be a pseudo-abnormal unit. In addition, the numerical value provided to the abnormal unit and the pseudo abnormal unit can be arbitrarily set.

  Here, the pseudo abnormal unit means that the difference from the luminance of the corresponding observation unit in the reference image is not equal to or greater than the abnormal unit determination threshold, but is equal to or greater than the pseudo abnormal unit determination threshold smaller than the abnormal unit determination threshold. It is an observation unit in a certain discrimination target image. That is, the pseudo-abnormal unit is an observation unit in which the difference from the luminance of the corresponding observation unit in the reference image is large to some extent but not so large as to be determined as an abnormal unit. Note that the value of the pseudo-abnormal unit determination threshold value can be arbitrarily set. The pseudo-abnormal unit is not used when creating an abnormal unit connected component. That is, the abnormal unit connected component does not include a pseudo abnormal unit.

  In the example shown in FIG. 11 (a-1), since the observation unit of the addresses 2-2, 3-2 and 4-2 indicated by xy is an abnormal unit, it has a numerical value 5 and the address is 4. Since the observation unit of -4 is a pseudo-abnormal unit, a numerical value 2 is provided. It is assumed that the observation unit whose numerical value is not shown has a numerical value of 0. Then, the post-determination processing unit 62 adds the abnormal unit determined by the determination processing unit 61 to the buffer memory. Here, as shown in FIG. 11 (b-1), the buffer memory includes the same number of observation units as the discrimination target image. In the initial state of the buffer memory, no data is stored, that is, all observation units do not contain numbers. It is assumed that the observation unit whose numerical value is not shown has a numerical value of 0. Then, when the abnormal unit determined by the determination processing unit 61 as shown in FIG. 11 (a-1) is added to the buffer memory, the buffer memory becomes as shown in FIG. 11 (b-1). . In this case, in the buffer memory in the initial state, all observation units do not contain numbers, so the buffer memory shown in FIG. 11 (b-1) with the abnormal units added is shown in FIG. 11 (a-1). It is a copy of the results shown.

  Subsequently, the post-discrimination processing unit 62 decrements (decreases) the numerical values of all observation units in the buffer memory shown in FIG. It should be noted that the numerical value of the observation unit having a numerical value of 0 is assumed to remain 0 even if it is decremented.

  Next, it is assumed that the result of the discrimination processing unit 61 discriminating the abnormal unit based on the second discrimination target image is FIG. 11A-2. In this case, only the observation unit whose address is 2-2 is an abnormal unit, is provided with a numerical value 5, and the numerical values of the other observation units are 0. Then, when the determination post-processing unit 62 adds the result of the determination of the abnormal unit by the determination processing unit 61 to the buffer memory, the address in the buffer memory is 2-2 as shown in FIG. A numerical value is added only to the observation unit. In the buffer memory, the numerical value of the observation unit does not exceed 5. That is, the maximum value of the observation unit is 5. Further, the numerical values of all the observation units are decremented by 1 from the state shown in FIG. Therefore, as shown in FIG. 11 (b-2), the observation unit whose address is 2-2 is 5, the observation unit whose addresses are 3-2 and 4-2 is 4, and the observation unit whose address is 4-4 is It is 1. Subsequently, the post-determination processing unit 62 decrements the numerical values of all the observation units in the buffer memory shown in FIG.

  Next, assume that the result of the discrimination processing unit 61 discriminating abnormal units based on the third discrimination target image is shown in FIG. In this case, only the observation unit whose address is 3-2 is an abnormal unit and has a numerical value 5, and only the observation unit whose address is 3-4 is a pseudo-abnormal unit and has a numerical value 2. Then, when the post-determination processing unit 62 adds the result of the determination of the abnormal unit by the determination processing unit 61 to the buffer memory, as shown in FIG. 11B-3, the address in the buffer memory is 3- Numeric values are added only to 2 and 3-4 observation units. In this case, the numerical values of all the observation units are decremented by 1 from the state shown in FIG. Therefore, as shown in FIG. 11B-3, the observation unit with the address 2-2 is 4, the observation unit with the address 3-2 is 5, the observation unit with the address 4-2 is 3, and the address is The observation unit of 3-4 is 2.

  Thereafter, the same operation is repeated until the molding is completed. Note that the same operation is repeated n times. When n reaches a predetermined number, the post-determination processing unit 62 may end the rewriting of the buffer memory. Here, the n can be arbitrarily set by inputting from the setting input unit 68 and may be any number, for example, may be 1 or may be infinite. Further, the post-determination processing unit 62 extracts an observation unit whose numerical value is greater than or equal to the threshold value for erroneous determination candidate in the buffer memory as an erroneous determination candidate region. Here, the threshold value for misidentification candidates can be arbitrarily set by inputting from the setting input unit 68, and may be any number. In addition, the operator can appropriately input a command to end the rewriting of the buffer memory from the setting input unit 68 as appropriate.

  In the present embodiment, the post-discrimination processing unit 62 extracts a misclassification candidate region every time the buffer memory is rewritten based on the result of the discrimination unit 61 discriminating an abnormal unit, and is stored in the misclassification candidate database 55. The erroneous determination candidate area is updated. For example, if the threshold value for misidentification candidate is 1, it is shown in FIG. 11B-1 which is rewritten by the result of the discrimination processing unit 61 discriminating abnormal units based on the first discrimination target image. Based on such a buffer memory, observation units having addresses 2-2, 3-2, 4-2, and 4-4 are extracted as misclassification candidate regions. The observation units having addresses 2-2, 3-2, 4-2, and 4-4 are stored in the misclassification candidate database 55 as misclassification candidate areas.

  Next, the address is 2 based on the buffer memory as shown in FIG. 11 (b-2) rewritten by the result of the determination of the abnormal unit by the determination processing unit 61 based on the second determination target image. -2, 3-2, 4-2, and 4-4 observation units are extracted as misclassification candidate regions. Then, the post-determination processing unit 62 uses the observation units with addresses 2-2, 3-2, 4-2, and 4-4 as new misdetermination candidate areas, and the misjudgment already stored in the misdetermination candidate database 55. Update another candidate area. In this case, since the new misclassification candidate area and the already stored misclassification candidate area are the same observation unit, there is no substantial change.

  Next, the address is 2 based on the buffer memory as shown in FIG. 11 (b-3) rewritten by the result of the determination of the abnormal unit by the determination processing unit 61 based on the third determination target image. -2, 3-2, 4-2, and 3-4 observation units are extracted as misclassification candidate regions. Then, the post-determination processing unit 62 sets the observation units whose addresses are 2-2, 3-2, 4-2, and 3-4 as new misdetermination candidate areas, and the misjudgment already stored in the misdetermination candidate database 55. Update another candidate area. As a result, the observation unit with the address 4-4 is not a misclassification candidate area, and the observation unit with the address 3-4 is newly registered as a misclassification candidate area.

  Thereafter, by repeating the same operation, the erroneous determination candidate areas stored in the erroneous determination candidate database 55 are sequentially updated. Thereby, misclassification learning can be performed while performing actual mold monitoring. Then, the discrimination processing unit 61 discriminates the presence / absence of an image abnormality for the discrimination target image excluding the erroneous discrimination candidate area stored in the erroneous discrimination candidate database 55 that is sequentially updated.

  Further, the display unit 66 can display the color so that the color of the observation unit changes according to the numerical value of the buffer memory that is sequentially rewritten. Thereby, the operator can easily grasp the misidentification candidate area extracted in response to the change of the misidentification candidate threshold. Therefore, based on experience, the operator selects a threshold value for the misclassification candidate so that only observation units corresponding to areas that are considered to be appropriate as areas that cause misclassification are extracted as misclassification candidate areas. Can be set.

  Thus, in the present embodiment, misclassification learning is performed while the injection molding machine 10 is performing actual molding, that is, while performing actual mold monitoring. For this reason, it is not necessary to perform misclassification learning before the molding process of the injection molding machine 10 is started, so that the throughput of the injection molding machine 10 can be improved. Further, since the misclassification candidates stored in the misclassification candidate database 55 are sequentially updated, even if the misclassification area changes due to an environment that changes with time, it can be dealt with sequentially.

  Next, a fourth embodiment of the present invention will be described. In addition, about the thing which has the same structure as 1st-3rd embodiment, the description is abbreviate | omitted by providing the same code | symbol. Explanation of the same operations and effects as those of the first to third embodiments is also omitted.

  FIG. 12 is a diagram showing a method for discriminating an abnormality of a discrimination target image on a mold surface of a movable mold according to the fourth embodiment of the present invention, and FIG. 13 is a movable mold according to the fourth embodiment of the present invention. It is a flowchart which shows the operation | movement which discriminate | determines abnormality of the discrimination | determination target image of the metal mold | die surface.

  As described in the first embodiment, when an abnormal unit connected component is created and the presence or absence of an image abnormality is determined based on whether or not the size of the abnormal unit connected component exceeds a component size determination threshold When the presence / absence of image abnormality is determined for the determination target image excluding the erroneous determination candidate area stored in the erroneous determination candidate database 55, the abnormal unit connected component is divided by the erroneous determination candidate area, and the size of the abnormal unit connected component is determined. May become smaller. For example, in the example shown in FIG. 12, if the observation unit C is a misidentification candidate region, the abnormal unit connected component B has a size of 3 when the abnormal unit connected component is obtained after the observation unit C is excluded. Divided into two anomalous unit connected components. In this case, the abnormality may be overlooked.

  Therefore, in the present embodiment, the discrimination processing unit 61 first discriminates the presence / absence of an image abnormality for a discrimination target image that does not exclude an erroneous discrimination candidate area. In this case, the determination processing unit 61 creates an abnormal unit connected component, and determines whether there is an image abnormality based on whether or not the size of the abnormal unit connected component exceeds a component size determination threshold. For example, in the example shown in FIG. 12, when the hatched observation unit is an abnormal unit, the size of the abnormal unit connected component is 5, 2, 3 and 7 from left to right. Here, it is assumed that the size of the abnormal unit connected component B is 7 and exceeds the component size determination threshold.

  When the size of the abnormal unit connected component exceeds the component size determination threshold, the determination processing unit 61 calculates an abnormal unit in the abnormal unit connected component B excluding the erroneous determination candidate region. In this case, the discrimination processing unit 61 corrects the size of the abnormal unit connected component for the discrimination target image that does not exclude the misjudgment candidate region and whose size exceeds the component size discrimination threshold by the misjudgment candidate region. The presence / absence of an image abnormality is determined based on whether or not the correction value of the size of the abnormal unit connected component exceeds the correction value of the component size determination threshold.

That is, the erroneous determination candidate region existing in the abnormal unit connected component from the size of the abnormal unit connected component targeted for the determination target image that does not exclude the erroneous determination candidate region and whose size exceeds the component size determination threshold A value obtained by subtracting the size of the observation unit as a correction value for the size of the abnormal unit connected component. A value obtained by multiplying the component size determination threshold value by a coefficient α of 0 or more and 1 or less is set as a correction value for the component size determination threshold value. Then, when the correction value of the size of the abnormal unit connected component exceeds the correction value of the component size determination threshold, it is determined that there is an image abnormality. In this case, it is determined that there is an image abnormality when the following expression (1) is satisfied.
(Size of an abnormal unit connected component when an erroneous determination candidate region is not excluded) − (Size of an erroneous determination candidate region existing in an abnormal unit connected component)> α × (Threshold for component size determination) Equation (1) )
If the size of the abnormal unit connected component targeted for the determination target image that does not exclude the erroneous determination candidate region does not exceed the component size determination threshold, the determination processing unit 61 determines that there is no image abnormality. Also, when the formula (1) is not established, the determination processing unit 61 determines that there is no image abnormality.

Next, a flowchart will be described.
Step S11 The timing signal is input to the timing signal input unit 64.
Step S12 The imaging device 41 captures an image of the mold surface including the cavity 13a of the movable mold 13.
Step S13 The discrimination processing unit 61 configures each pixel as an observation unit.
Step S14: A determination is made for each observation unit.
Step S15: An abnormal unit connected component size is created.
Step S16: It is determined whether or not the abnormal unit connected component size (when the erroneous determination candidate region is not excluded) exceeds a component size determination threshold. If the abnormal unit connected component size exceeds the component size determination threshold, the process proceeds to step S17. If the abnormal unit connected component size does not exceed the component size determination threshold, the process proceeds to step S19.
Step S17: It is determined whether or not the correction value for the abnormal unit connected component size exceeds the correction value for the component size determination threshold. If the correction value of the abnormal unit connected component size exceeds the correction value of the component size determination threshold value, the process proceeds to step S18, and if the correction value of the abnormal unit connected component size does not exceed the correction value of the component size determination threshold value, step Proceed to S19.
Step S18: It is determined that there is an image abnormality, and the process is terminated.
Step S19: It is determined that there is no image abnormality, and the process is terminated.

  As described above, in the present embodiment, the determination processing unit 61 first determines the presence / absence of an image abnormality for a determination target image that does not exclude the erroneous determination candidate region. The presence / absence of an image abnormality is determined based on whether or not the correction value of the component size exceeds the correction value of the component size determination threshold value. Therefore, even if the abnormal unit connected component is divided by the erroneous determination candidate region, the abnormality is not overlooked.

  In the first to fourth embodiments, the horizontal type injection molding machine in which the movable platen moves in the horizontal direction (horizontal direction) has been described. However, in the present invention, the movable platen is in the vertical direction (vertical direction). It can also be applied to a vertical type injection molding machine that moves to). Also, for mold monitoring, confirm that there is a molded product on the mold surface of the movable mold at the time of mold opening, and confirm that the molded product has been removed on the mold surface of the movable mold However, the present invention can also be applied to confirming the quality and shape of the molded product on the mold surface of the movable mold at the time of mold opening. Furthermore, the present invention can be applied to molding machines such as die casting machines and IJ sealing presses in addition to injection molding machines.

  The present invention is not limited to the above-described embodiment, and various modifications can be made based on the spirit of the present invention, and they are not excluded from the scope of the present invention.

It is a block diagram which shows the structure of the control apparatus of the metal mold | die monitoring apparatus in the 1st Embodiment of this invention. It is the schematic of the molding machine in the 1st Embodiment of this invention. It is a block diagram which shows the apparatus structure of the metal mold | die monitoring apparatus applied to the molding machine in the 1st Embodiment of this invention. It is a figure which shows the state which is a metal mold | die surface of the movable metal mold | die in the 1st Embodiment of this invention, and the molded article has adhered. It is a figure which is the metal mold | die surface of the movable metal mold | die in the 1st Embodiment of this invention, and shows the state which released the molded article. It is a figure which shows the brightness | luminance of the area | region of the A section of FIG. 4 in the 1st Embodiment of this invention. It is a figure which shows the method of calculating the size of the abnormal unit connection component of the discrimination | determination target image of the metal mold | die surface of a movable mold in the 1st Embodiment of this invention. It is a flowchart which shows the operation | movement which discriminate | determines abnormality of the discrimination | determination object image of the metal mold | die surface of a movable metal mold | die in the 1st Embodiment of this invention. It is a figure which shows the method of putting together the pixel of the discrimination | determination target image of the metal mold | die surface of a movable mold in the 1st Embodiment of this invention, and comprising an observation unit. It is a figure which shows the change of the buffer of the discrimination target image of the mold surface of the movable mold in the first embodiment of the present invention. It is a figure which shows the change of the buffer of the discrimination target image of the mold surface of the movable mold in the third embodiment of the present invention. It is a figure which shows the method to discriminate | determine abnormality of the discrimination | determination object image of the metal mold | die surface of a movable metal mold | die in the 4th Embodiment of this invention. It is a flowchart which shows the operation | movement which discriminate | determines abnormality of the discrimination | determination object image of the metal mold | die surface of a movable metal mold | die in the 4th Embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Injection molding machine 12 Fixed mold 13 Movable mold 52 Reference | standard image database 55 Misidentification candidate database 61 Discrimination processing part 62 Discrimination post-processing part 67 Image input part

Claims (11)

(A) an image input unit for acquiring an image of a mold surface of a movable mold or a fixed mold of a molding machine;
(B) a reference image database for storing a reference image acquired in advance by the image input unit;
(C) a determination processing unit that compares the determination target image acquired by the image input unit with the reference image, executes a determination process, and determines whether the determination target image is abnormal;
(D) a misclassification candidate database storing a misclassification candidate area extracted by incrementing or decrementing the characteristic value of the observation unit in the discrimination target image with respect to the characteristic value of the observation unit in the previous discrimination target image. With
(E) The determination processing unit compares the characteristic value of the observation unit in the determination target image excluding the erroneous determination candidate region with the characteristic value of the observation unit in the reference image, and compares the observation unit in the determination target image. It is determined whether it is an abnormal unit every time, and it is determined that there is an abnormality in the determination target image when the number of the abnormal units or the size of the abnormal unit connected component combining the abnormal units exceeds a threshold value. A mold monitoring apparatus characterized by: The mold monitoring apparatus according to claim 1 , wherein the determination processing unit corrects the size of the abnormal unit connected component based on the erroneous determination candidate region. The mold monitoring apparatus according to claim 1 , wherein the determination processing unit determines whether there is an abnormality in the determination target image by excluding the erroneous determination candidate region from the determination target image. Mold apparatus according to any one of claim 1 3, having a discrimination post-processing unit for extracting the misidentified candidate region based on a result of the determination process of the determination processing section. The mold monitoring apparatus according to claim 4 , wherein the post-determination processing unit stores the extracted misdetermination candidate region in the misdetermination candidate database. The post-determination processing unit extracts the misdetermination candidate region based on the result of the determination process of the determination processing unit executed in misdetermination learning in which an operation that causes misdetermination before molding of the molding machine is performed. The mold monitoring apparatus according to claim 4 or 5 . The mold monitoring apparatus according to claim 4 or 5 , wherein the post-discrimination processing unit extracts the erroneous discrimination candidate area based on a result of the discrimination processing of the discrimination processing unit executed during molding of the molding machine. (A) The image input unit acquires an image of the mold surface of the movable mold or the fixed mold of the molding machine,
(B) storing a reference image acquired in advance by the image input unit in a reference image database;
(C) The discrimination processing unit performs discrimination processing by comparing the discrimination target image acquired by the image input unit with a reference image acquired in advance, and determines whether the discrimination target image is abnormal. ,
(D) discriminating the post-processing unit extracts misclassification candidate region by incrementing or decrementing the characteristic values of the observation unit in the determination target image for the characteristic values of the observation unit in the previous determination target image, the erroneous While storing another candidate area in the misclassification candidate database ,
(E) The determination processing unit compares the characteristic value of the observation unit in the determination target image excluding the erroneous determination candidate region with the characteristic value of the observation unit in the reference image, and compares the observation unit in the determination target image. It is determined whether it is an abnormal unit every time, and it is determined that there is an abnormality in the determination target image when the number of the abnormal units or the size of the abnormal unit connected component combining the abnormal units exceeds a threshold value. A mold monitoring method characterized by: The mold monitoring method according to claim 8 , wherein when the discrimination target image is acquired, the discrimination target image is acquired by changing the surrounding environment while the mold is stationary. The mold monitoring method according to claim 8 , wherein when the discrimination target image is acquired, the discrimination target image is acquired by changing a position of a movable portion of the mold while the mold is stationary. (A) Computer for mold monitoring
(B) an image input unit for acquiring an image of a movable mold or a fixed mold of the molding machine;
(C) a reference image database for storing a reference image acquired in advance by the image input unit;
(D) a determination processing unit that compares the determination target image acquired by the image input unit with the reference image, executes a determination process, and determines whether or not the determination target image is abnormal;
(E) It is made to function as a misjudgment candidate database which stores the misjudgment candidate area extracted by incrementing or decrementing the characteristic value of the observation unit in the discrimination target image with respect to the characteristic value of the observation unit in the previous discrimination target image. With
(F) The discrimination processing unit compares the characteristic value of the observation unit in the discrimination target image excluding the erroneous discrimination candidate area with the characteristic value of the observation unit in the reference image, and compares the observation unit in the discrimination target image. It is determined whether it is an abnormal unit every time, and it is determined that there is an abnormality in the determination target image when the number of the abnormal units or the size of the abnormal unit connected component combining the abnormal units exceeds a threshold value. mold monitoring program to be.

Images