Inspection device, injection molding system, and inspection method

Abstract

To provide an inspection device, an injection molding system and an inspection method, capable of determining quality of a molding with a small load.SOLUTION: An inspection device includes an imaging device that acquires a polarization image of a molding, and a determination unit that determines quality of the molding. The determination unit includes an image generation unit that generates a pseudo image from an input image according to a learning model subjected to machine learning. The polarization image or an arithmetic image obtained by calculation from the polarization image is input to the image generation unit as an input image, and quality of the molding is determined on the basis of the input image and the pseudo image generated by the image generation unit.SELECTED DRAWING: Figure 9

Description

[Technical Field]

[0001] The present invention relates to an inspection device, an injection molding system, and an inspection method. [Background technology]

[0002] Patent Document 1 shows a system having a light source that illuminates an injection-molded product through a polarizing plate, an imaging means that images the injection-molded product through the polarizing plate, and a setting value correction means that corrects the setting values ​​of the molding process based on the polarization stripe pattern obtained by the imaging means. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 1-120317 Summary of the Invention [Problem to be solved by the invention]

[0004] It would be preferable to be able to determine the quality of an injection-molded product with a small load. An object of the present invention is to provide an inspection device, an injection molding system, and an inspection method that can determine the quality of a molded product with a small load. [Means for solving the problem]

[0005] The inspection device according to the present invention comprises: Molded products Multiple polarization channels an imaging device for acquiring a polarized image; a judgment unit for judging whether the molded product is good or bad; Equipped with The determination unit an image generation unit that generates a pseudo image from an input image in accordance with a machine-learned learning model; The polarized image or a calculated image obtained by calculation from the polarized image is input to the image generation unit as the input image, and the quality of the molded product is determined based on the input image and the pseudo image generated by the image generation unit. death, the image generation unit generates pseudo images having the same number of polarization channels from the polarization images or the calculated images of the plurality of polarization channels; The determining unit compares the polarized image or the calculated image with the pseudo image for the same polarization channel and determines whether the image is good or bad based on the result of the comparison. do.

[0006] The injection molding system according to the present invention comprises: The apparatus includes an injection molding machine and the above-described inspection device for inspecting molded products molded by the injection molding machine.

[0007] The inspection method according to the present invention comprises: Imaging device for molded products Multiple polarization channels Obtain a polarized image, inputting the polarized image or a calculated image obtained by calculation from the polarized image to an image generating unit as an input image; The molded product is judged to be good or bad based on the pseudo image of the input image generated by the image generation unit and the input image. death, the image generation unit generates pseudo images having the same number of polarization channels from the polarization images or the calculated images of the plurality of polarization channels; The polarized image or the calculated image is compared with the pseudo image between the same polarization channels, and a pass / fail judgment is made based on the result of the comparison. do. [Effects of the Invention]

[0008] According to the present invention, it is possible to provide an inspection device, an injection molding system, and an inspection method that can determine the quality of a molded product with a small load. [Brief explanation of the drawings]

[0009] [Figure 1] 1 is a diagram showing an example of a schematic configuration of an injection molding system according to an embodiment; [Figure 2] FIG. 2 is an example of a block diagram illustrating functions of a control device and a processing device. [Figure 3] FIG. 1 is a diagram illustrating an example of a schematic configuration of an imaging device. [Figure 4] FIG. 2 is a diagram illustrating an example of a polarized image acquired by an imaging device. [Figure 5] FIG. 10 is a diagram illustrating an example of a schematic configuration of an imaging device according to a first modification. [Figure 6] FIG. 10 is a diagram illustrating an example of a schematic configuration of an imaging device according to a second modification. [Figure 7] FIG. 11 is a diagram illustrating an example of a schematic configuration of an imaging device according to a third modification. [Figure 8] 10 is a flowchart showing an example of a procedure of an inspection process performed by the processing device. [Figure 9] FIG. 3 is a diagram illustrating an inspection process according to the first embodiment. [Figure 10] FIG. 10 is a diagram illustrating an inspection process according to a second embodiment. [Figure 11] FIG. 10 is a diagram illustrating an inspection process according to a third embodiment. [Figure 12] FIG. 10 is a diagram illustrating an inspection process according to a fourth embodiment. [Figure 13] FIG. 13 is a diagram illustrating an inspection process according to a fifth embodiment. [Figure 14] FIG. 10 is an image diagram showing an example of display of inspection results. [Figure 15] 10 is a flowchart illustrating an example of a procedure of machine learning processing executed by the inspection device. [Figure 16] FIG. 10A is a diagram illustrating the machine learning processing of the first to fourth embodiments, and FIG. 10B is a diagram illustrating the machine learning processing of the fifth embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

[0011] Fig. 1 is a diagram showing an example of a schematic configuration of an injection molding system 1 according to the first embodiment. Fig. 2 is an example of a block diagram showing the functions of a control device 40 and a processing device 120.

[0012] The injection molding system 1 includes an injection molding machine 2 and an inspection device 100 that inspects the quality (normality or abnormality) of a molded product molded by the injection molding machine 2.

[0013] 《Injection molding machine 2》 First, a description will be given of the injection molding machine 2. In the following description, the direction in which resin is injected will be referred to as the front side, and the direction opposite to the direction in which resin is injected will be referred to as the rear side.

[0014] The injection molding machine 2 includes a mold clamping device (not shown), an injection device 10, a material supply device 81, a control device 40 that controls the entire device, an operation unit 51 that accepts user input operations, and a display unit 52 that displays operation acceptance screens and images.

[0015] The mold clamping device, injection device 10, material supply device 81, and control device 40 will be described in detail later.

[0016] The operation unit 51 can be exemplified by an input device such as a button, a switch, or a touch panel. The display unit 52 can be exemplified by a liquid crystal display or an organic EL display. The operation unit 51 and the display unit 52 may be integrally configured.

[0017] The injection molding machine 2 repeatedly produces molded products through a cycle consisting of a mold closing process, mold clamping process, filling process, pressure holding process, cooling process, metering process, mold opening process, and ejection process. The mold closing process is a process of closing a mold device consisting of a fixed mold and a movable mold. The mold clamping process is a process of tightening the mold device. The filling process is a process of pouring molten resin into the mold device. The pressure holding process is a process of applying pressure to the poured resin. The cooling process is a process of solidifying the resin in the mold device after the pressure holding process. The metering process is a process of metering the molten resin for the next molded product. The mold opening process is a process of opening the mold device. The ejection process is a process of ejecting the molded product from the mold device after the mold is opened. Note that, to shorten the molding cycle, the metering process may be performed while the cooling process is being carried out.

[0018] (mold clamping device) The mold clamping device includes a fixed platen on which a fixed mold is attached and a movable platen on which a movable mold is attached, and performs mold closing, mold clamping, and mold opening by moving the movable platen forward and backward to move the movable mold toward and away from the fixed mold. The type of mold clamping device is not particularly limited. Examples include a toggle type using an electric motor and a toggle mechanism, a direct pressure type using a fluid pressure cylinder, and an electromagnetic type using a linear motor and an electromagnet.

[0019] (Injection device 10) The injection device 10 has a cylinder 11 for heating resin as a molding material, and a nozzle 12 disposed at the front end of the cylinder 11. The injection device 10 also has a screw 20 disposed within the cylinder 11 that is rotatable and can move back and forth along the rotation axis, heaters h11, h12, and h13 as a heat source for heating the cylinder 11, and a drive device 60 disposed behind the cylinder 11.

[0020] The screw 20 has a screw body 21 and an injection section 22 disposed forward of the screw body 21, and is connected to a drive unit 60 via a shaft portion at the rear end. The screw body 21 has a flight section 23 and a pressure member 24 disposed detachably at the front end of the flight section 23. The flight section 23 has a rod-shaped main body section 23a and a spiral flight 23b formed to protrude from the outer peripheral surface of the main body section 23a, and a spiral thread groove 26 is formed along the flight 23b. From the rear end to the front end of the flight section 23, the depth of the thread groove 26 is constant, which can be exemplified as a constant screw compression ratio.

[0021] The screw 20 may have a flight portion 23 formed over the entire screw body 21 without having a pressure member 24. The screw body 21 may also be divided, from its rear end to its front end, into a supply portion where resin is supplied, a compression portion where the supplied resin is melted while being compressed, and a metering portion where the molten resin is metered in fixed amounts. The depth of the screw groove 26 is preferably deepest in the supply portion and shallowest in the metering portion, and becomes shallower from the rear side to the front side in the compression portion.

[0022] The injection section 22 has a head section 31 with a conical portion at the tip, a rod section 32 formed adjacent to the rear side of the head section 31, a check ring 33 disposed around the rod section 32, and a seal ring 34 attached to the front end of the pressure member 24.

[0023] During the metering process, as the screw 20 moves backward, the check ring 33 is moved forward relative to the rod portion 32, and when it is separated from the seal ring 34, resin is sent from the rear side to the front side of the injection portion 22. Also, during the injection process, as the screw 20 moves forward, the check ring 33 is moved rearward relative to the rod portion 32, and when it is brought into contact with the seal ring 34, backflow of resin is prevented.

[0024] A resin supply port 14 serving as a molding material supply port is formed at the rear of the cylinder 11. The resin supply port 14 is formed at a location facing the rear end of the thread groove 26 when the screw 20 is positioned at the frontmost position within the cylinder 11. A material supply device 81 that supplies resin into the cylinder 11 is attached to the resin supply port 14.

[0025] The drive device 60 is a device that rotates and moves the screw 20 forward and backward within the cylinder 11. The drive device 60 has a metering motor 61 as a drive source that rotates the screw 20 within the cylinder 11, and an injection motor 71 as a drive source that moves the screw 20 in the direction of the rotation axis within the cylinder 11. The metering motor 61 and the injection motor 71 can be, for example, servo motors.

[0026] A motion conversion mechanism or the like is provided between the injection motor 71 and the screw 20 to convert the rotational motion of the injection motor 71 into linear motion of the screw 20. For example, the motion conversion mechanism has a screw shaft and a screw nut that screws onto the screw shaft. For example, balls, rollers, or the like may be provided between the screw shaft and the screw nut. The drive source that moves the screw 20 in the rotational axis direction is not limited to the injection motor 71, and may be, for example, a hydraulic cylinder or the like.

[0027] (Material supply device 81) The material supply device 81 has a hopper 82 that stores molding material (e.g., resin pellets), a feed cylinder 83 that extends horizontally from the bottom end of the hopper 82, and a cylindrical guide part 84 that extends downward from the front end of the feed cylinder 83. The material supply device 81 also has a feed screw 85 that is rotatably disposed within the feed cylinder 83, and a feed motor 86 that rotates the feed screw 85.

[0028] Resin supplied from inside the hopper 82 into the feed cylinder 83 is advanced along the thread groove of the feed screw 85 as the feed screw 85 rotates. Resin sent from the front end of the feed screw 85 into the guide section 84 falls inside the guide section 84 and is supplied into the cylinder 11.

[0029] The feed cylinder 83 does not necessarily have to extend horizontally, but may extend obliquely relative to the horizontal, and the outlet side of the feed cylinder 83 may be higher than the inlet side.

[0030] The resin supplied into the feed cylinder 83 may be heated by a heater (not shown). At this time, the resin is preferably heated to a temperature at which it does not melt, for example, a predetermined temperature below the glass transition point.

[0031] (Control device 40) The control device 40 has a CPU 41, a ROM 42 that stores control programs and the like, a readable and writable RAM 43 that stores calculation results and the like, a storage unit 44 such as a hard disk, an input interface (I / F) 45, and an output interface (I / F) 46. The control device 40 realizes various functions by causing the CPU 41 to execute programs stored in the ROM 42 or the storage unit 44 or the like.

[0032] The control device 40 may have a motor control unit 47 that controls the driving of the metering motor 61, injection motor 71, feed motor 86, etc., a heater control unit 48 that controls the temperature of heaters h11 to h13, and a parameter correction unit 49 that corrects the molding parameters when molding the molded product 3.

[0033] (Operation of injection molding machine 2) The operation of the injection molding machine 2 controlled by the control device 40 will be described below. In the metering process, the motor control unit 47 of the control device 40 drives the metering motor 61 to rotate the screw 20. At this time, the motor control unit 47 drives the feed motor 86 to rotate the feed screw 85. An example of the motor control unit 47 is to rotate the screw 20 and the feed screw 85 synchronously during molding. The motor control unit 47 controls the current supplied to the metering motor 61 so that the rotational speed of the screw 20 becomes the rotational speed set via, for example, the operation unit 51. In addition, the motor control unit 47 controls the current supplied to the feed motor 86 so that the rotational speed of the feed screw 85 becomes the rotational speed set via, for example, the operation unit 51.

[0034] The resin supplied into the cylinder 11 by the material supply device 81 does not accumulate at the resin supply port 14, but is immediately sent to the front by the screw 20. The resin is not densely filled in the screw groove 26 of the screw 20, and the resin in the screw groove 26 is in a sparse state. Therefore, the faster the resin supply speed by the material supply device 81, the greater the amount of resin sent to the front by the screw 20 per unit time.

[0035] The resin supplied into the cylinder 11 is advanced along the thread groove 26 of the screw 20 as the screw 20 rotates, and is heated and melted by the heaters h11 to h13. The heater control section 48 of the control device 40 controls the power supplied to the heaters h11 to h13 so that the temperatures of the heaters h11 to h13 become the temperatures set via the operation section 51, for example.

[0036] Furthermore, the resin supplied into the cylinder 11 is gradually pressurized from the resin pressure increase start position in the screw body 21 to the front end of the screw body 21. The pressure increase start position is located a predetermined distance rearward from the pressure member 24, and is displaced according to the ratio (synchronization rate) between the rotation speed of the screw 20 and the rotation speed of the feed screw 85, etc. If the pressure increase start position is located within a predetermined distance from the pressure member 24, the molten state of the resin will stabilize, and the weight of the molded product will be stabilized.

[0037] The resin advanced along the screw groove 26 of the screw 20 passes through the resin flow path between the pressure member 24 and the cylinder 11, is kneaded during this period, and then is advanced through the resin flow path between the cylinder 11 and the rod portion 32. The resin is then sent to the front side of the screw 20 and accumulated in the front part of the cylinder. As the molten resin accumulates in front of the screw 20, the screw 20 moves backward.

[0038] In the metering process, the motor control unit 47 of the control device 40 controls the current supplied to the injection motor 71 so that the back pressure of the screw 20 becomes the back pressure set via, for example, the operation unit 51. By applying back pressure to the screw 20, the screw 20 is prevented from suddenly retracting, improving the kneading properties of the resin and making it easier for gas in the resin to escape to the rear side.

[0039] The motor control unit 47 monitors the position of the screw 20 using a position sensor (not shown) while the screw 20 is being retracted. When the screw 20 has retracted to the metering completion position and a predetermined amount of resin has accumulated in front of the screw 20, the control device 40 stops driving the metering motor 61. This stops the rotation of the screw 20 and completes the metering process. For example, the motor control unit 47 may stop driving the feed motor 86 and stop the rotation of the feed screw 85 simultaneously with the completion of the metering process.

[0040] In the filling step, the motor control unit 47 of the control device 40 drives the injection motor 71 to move the screw 20 forward and force the resin into the cavity space inside the mold device in a clamped state. At this time, the motor control unit 47 controls the current supplied to the injection motor 71 so that the movement speed of the screw 20 in the direction of the rotation axis becomes the movement speed set via the operation unit 51, for example.

[0041] In the pressure holding process, the motor control unit 47 controls the current supplied to the injection motor 71 so that the resin pressure becomes the pressure set via, for example, the operation unit 51. As a result, the resin filled in the cavity space shrinks due to cooling, but the resin to compensate for the shrinkage is replenished.

[0042] The set values ​​of rotation speed, movement speed, pressure, etc. used by the motor control unit 47 when controlling various motors, and the set values ​​of temperature used by the heater control unit 48 when controlling the heaters h11 to h13 are stored as molding parameters in the ROM 42 or the memory unit 44, etc.

[0043] The processing device 120 of the inspection device 100 sends a pass / fail judgment of the molded product to the parameter correction unit 49. The parameter correction unit 49 uses the judgment result that the molded product 3 is defective output from the processing device 120 and the molding parameters used when the molded product 3 was molded in the injection molding machine 2 to correct the molding parameters to be used when molding the molded product 3 from the next time onwards.

[0044] When the processing device 120 outputs a determination result indicating that the molded product 3 is a defective product, the parameter correction unit 49 determines that the molding parameters used to mold the molded product 3 in the injection molding machine 2 were not good. Examples of molding parameters include the movement speed of the screw 20 in the filling process and the temperatures of the heaters h11 to h13. For example, wear on the mold device for the molded product 3 can cause a deterioration in the fluidity of the molten resin within the mold device, resulting in a change in stress distribution. Therefore, an example of the parameter correction unit 49 is to change at least one of the movement speed of the screw 20 and the temperatures of the heaters h11 to h13. This is because an increase in the movement speed of the screw 20 improves the fluidity of the molten resin within the mold device, and an increase in the temperatures of the heaters h11 to h13 improves the fluidity of the molten resin within the mold device.

[0045] Inspection device 100 As shown in FIG. 1, the inspection device 100 includes an imaging device 110 that captures an image of the molded product 3, and a processing device 120 that processes the image output from the imaging device 110.

[0046] (imaging device 110) Fig. 3 is a diagram showing an example of a schematic configuration of the imaging device 110. Fig. 4 is a diagram showing an example of a polarized image acquired by the imaging device 110.

[0047] The imaging device 110 captures a molded product 3 molded by an injection molding machine 2 as an image capture target, and acquires a polarized image of the molded product 3 containing multiple polarization channels. A polarized image refers to an image of the image capture target captured by refraction and reflection of polarized light. If images of the same image capture target are captured using different types of polarization light at the same angle, direction, and distance, different images of the same object will be obtained. In other words, by capturing images using multiple types of polarization, multiple component images of the same object can be obtained. A polarization channel refers to an image of one of the multiple component images.

[0048] The imaging device 110 includes a light source 111 that generates light, a linear polarizer 112 that creates linearly polarized light from the light emitted from the light source 111, a wavelength plate 113 that converts the linearly polarized light created by the linear polarizer 112 into circularly polarized light, and a polarization camera 114.

[0049] Examples of light source 111 include lighting such as a lamp, an incandescent bulb, a fluorescent lamp, and an LED. The light emitted from light source 111 is not limited to visible light, and may be infrared light. The light preferably has a wavelength of 360 to 900 nm.

[0050] The linear polarizer 112 is an optical element that produces linearly polarized light from the light emitted from the light source 111 .

[0051] The wave plate 113 can be exemplified as a quarter wave plate that produces a phase difference of 90 degrees.

[0052] Polarization camera 114 can be exemplified as a camera in which polarizers at 0 degrees, 45 degrees, 90 degrees, and 135 degrees are regularly arranged between the imaging element and the lens, and which can acquire a polarization image containing four polarization channels corresponding to the above four polarization angles in a single image capture. Polarization camera 114 may also be a camera that can generate an image by calculating the polarization direction and polarization degree using the above polarization image (for example, by performing arithmetic operations, trigonometric functions, or inverse trigonometric functions).

[0053] In the imaging device 110 configured as above, the molded product 3 molded by the injection molding machine 2 is placed above the wavelength plate 113, and the light transmitted through the molded product 3 is imaged by the polarization camera 114.

[0054] As shown in Figure 4, the above imaging method allows the acquisition of polarization images in four polarization channels corresponding to four different polarization angles.

[0055] In the imaging device 110, linearly polarized light that has passed through the linear polarizer 112 may be incident on the molded article 3 without using the wave plate 113. However, when linearly polarized light is incident on the molded article 3, if the main axis direction of the molded article 3 and the polarization direction are perpendicular to each other, the light will not be able to pass through the molded article 3, which may result in loss of information. Therefore, it is preferable to use the wave plate 113 to incident circularly polarized light on the molded article 3. By incident circularly polarized light on the molded article 3, loss of information can be suppressed.

[0056] The wave plate 113 may also be a half-wave plate. If a half-wave plate with an optical axis tilted at an azimuth angle of 45 degrees is used as the wave plate 113, the polarization direction can be rotated vertically. This allows the polarization direction of the linearly polarized light that has passed through the linear polarizer 112 to be changed and incident on the molded article 3, thereby preventing information loss. In other words, if the polarization direction is perpendicular to the main axis direction of the molded article 3 and light cannot pass through the molded article 3, a fringe image corresponding to the stress distribution to be measured cannot be output. Therefore, by using a half-wave plate to vertically invert the polarization direction, information loss due to light not being able to pass through the molded article 3 can be prevented.

[0057] The wave plate 113 may be an isolating plate or the like that converts linearly polarized light into elliptically polarized light.

[0058] (Modification of polarization camera) Fig. 5 is a diagram showing an example of a schematic configuration of an imaging device 110 according to Modification 1. Fig. 6 is a diagram showing an example of a schematic configuration of an imaging device 110 according to Modification 2. Fig. 7 is a diagram showing an example of a schematic configuration of an imaging device 110 according to Modification 3. Next, modifications of the imaging device 110 described above will be described. Elements similar to those in the imaging device 110 described above will be assigned the same reference numerals, and detailed descriptions thereof will be omitted.

[0059] (Variation 1) 5, imaging device 110 of modified example 1 includes light source 111, linear polarizer 112, polarization camera 114, and rotation means 215 that rotates linear polarizer 112. Rotation means 215 rotates film-like linear polarizer 112 by 45 degrees, 90 degrees, or 135 degrees around a line perpendicular to the plate surface as the center of rotation.

[0060] In the imaging device 110 configured as described above, a molded article 3 molded by an injection molding machine 2 is placed above the linear polarizer 112, and light transmitted through the molded article 3 is imaged by the polarization camera 114. Furthermore, the rotation means 215 rotates the linear polarizer 112 by 45 degrees, 90 degrees, and 135 degrees, and the polarization camera 114 images the light transmitted through the molded article 3 at each rotation angle. The polarization camera 114 acquires images corresponding to the four polarization angles at rotation angles of 0 degrees, 45 degrees, 90 degrees, and 135 degrees, as well as images obtained by calculating the polarization direction and polarization degree using these images. The configuration and rotation method of the rotation means 215 that rotates the linear polarizer 112 are not particularly limited. The linear polarizer 112 may be rotated by a robot or by hand.

[0061] Even with the imaging device 110 configured in this way, it is possible to obtain polarized images (FIG. 4) of four polarization channels corresponding to four different polarization angles.

[0062] (Variation 2) 6, the imaging device 110 of the second modification includes a light source 111, a linear polarizer 112, a polarization camera 114, and a rotation means 315 for rotating the molded article 3. The rotation means 315 rotates the molded article 3 by 45 degrees, 90 degrees, and 135 degrees around a line perpendicular to the plate surface of the film-like linear polarizer 112 as the rotation center.

[0063] In the imaging device 110 configured as described above, the molded article 3 molded by the injection molding machine 2 is placed above the linear polarizer 112, and the light transmitted through the molded article 3 is imaged by the polarization camera 114. Furthermore, the rotation means 315 rotates the molded article 3 by 45 degrees, 90 degrees, and 135 degrees, and the light transmitted through the molded article 3 at each rotation angle is imaged by the polarization camera 114. The polarization camera 114 acquires images corresponding to the four polarization angles at rotation angles of 0 degrees, 45 degrees, 90 degrees, and 135 degrees, as well as images in which the polarization direction and polarization degree are calculated using these images. The configuration and method of the rotation means 315 for rotating the molded article 3 are not particularly limited. The molded article 3 may be rotated by a robot or by hand.

[0064] Even with the imaging device 110 configured in this way, it is possible to obtain polarized images (FIG. 4) of four polarization channels corresponding to four different polarization angles.

[0065] (Variation 3) 7 , the imaging device 110 according to the third modification includes a light source 111, a linear polarizer 112, and a wave plate 113. The imaging device 110 also includes a first beam splitter 421 that reflects a portion of the light emitted from the light source 111 and transmits a portion of the light, a second beam splitter 422 that reflects a portion of the light that has transmitted through the first beam splitter 421 and transmits a portion of the light, and a third beam splitter 423 that reflects a portion of the light that has transmitted through the second beam splitter 422 and transmits a portion of the light. The imaging device 110 also includes a first linear polarizer 431 that creates linearly polarized light from the light reflected by the first beam splitter 421, and a first camera 441 that captures an image of the light that has transmitted through the first linear polarizer 431. The imaging device 110 also has a second linear polarizer 432 that creates linearly polarized light from the light reflected by the second beam splitter 422, and a second camera 442 that captures the light that has passed through the second linear polarizer 432. The imaging device 110 also has a third linear polarizer 433 that creates linearly polarized light from the light reflected by the third beam splitter 423, and a third camera 443 that captures the light that has passed through the third linear polarizer 433. The imaging device 110 also has a fourth linear polarizer 434 that creates linearly polarized light from the light that has passed through the third beam splitter 423, and a fourth camera 444 that captures the light that has passed through the fourth linear polarizer 434.

[0066] First camera 441 to fourth camera 444 have an imaging element and a lens, but unlike polarization camera 114, they are general cameras that do not have polarizers at 0 degrees, 45 degrees, 90 degrees, and 135 degrees.

[0067] The polarization axis (transmission axis) of the second linear polarizer 432 is inclined at 45 degrees with respect to the polarization axis of the first linear polarizer 431. The polarization axis (transmission axis) of the third linear polarizer 433 is inclined at 90 degrees with respect to the polarization axis of the first linear polarizer 431. The polarization axis (transmission axis) of the fourth linear polarizer 434 is inclined at 135 degrees with respect to the polarization axis of the first linear polarizer 431.

[0068] In the imaging device 110 configured as described above, the molded product 3 molded by the injection molding machine 2 is placed above the wavelength plate 113, and the light transmitted through the molded product 3 is captured by the first camera 441 to the fourth camera 444.

[0069] Even with the imaging device 110 configured in this way, it is possible to obtain polarized images (FIG. 4) of four polarization channels corresponding to four different polarization angles.

[0070] (Processing device 120) 1, the processing device 120 has a CPU 121, a ROM 122 that stores a control program and the like, a readable and writable RAM 123 that stores calculation results and the like, a storage unit 124 such as a hard disk, an input interface (I / F) 125, and an output interface (I / F) 126. The processing device 120 realizes various functions by causing the CPU 121 to execute programs stored in the ROM 122, the storage unit 124, or the like.

[0071] As shown in Figure 2, the processing device 120 has a receiving unit 131 that receives the polarized image of the molded product 3 output from the imaging device 110, a judgment unit 132 that judges whether the molded product 3 is good or bad using the polarized image of the molded product 3 received by the receiving unit 131, and an output unit 133 that outputs the result of the judgment made by the judgment unit 132 to the control device 40.

[0072] FIG. 8 is a flowchart showing an example of the procedure of the inspection process performed by the processing device 120. The processing device 120 repeatedly executes this process, for example, at a predetermined control period (for example, every second).

[0073] The processing device 120 determines whether the receiving unit 131 has received a polarized image of the molded article 3 from the imaging device 110 (S501). If a polarized image has not been received (NO in S501), the processing device 120 terminates the inspection process. On the other hand, if a polarized image has been received (YES in S501), the processing device 120 inputs the polarized image received in S501 or a calculated image calculated from the polarized image to the image generating unit 132a to generate a pseudo image (S502). The processing device 120 then calculates the difference between the polarized image or the calculated image and the pseudo image (S503). The processing device 120 then determines whether the difference calculated in S503 is equal to or greater than a predetermined threshold (S504). If the difference is equal to or greater than the threshold (YES in S504), the processing device 120 determines that the molded article 3 is abnormal (defective) (S505). On the other hand, if the difference is less than the threshold (NO in S504), the processing device 120 determines that the molded article 3 is normal (good) (S506). Thereafter, the processing device 120 outputs the determination result as to whether the molded article 3 is good or defective to the control device 40 (S507). The processes of S502, S503, S504, S505, and S506 are performed by the determination unit 132, and the process of S507 is performed by the output unit 133.

[0074] 《Judgment section 132》 Next, the configuration of the determination unit 132 and details of the inspection process executed in steps S502 to S507 in Fig. 8 will be described for the first to fifth embodiments. Figs. 9 to 13 are diagrams for explaining the determination unit of the first to fifth embodiments, respectively.

[0075] (Determination unit of the first embodiment) 9, the determination unit 132 has an image generation unit 132a that generates a pseudo image from an input image in accordance with a machine-learned learning model. The determination unit 132 further has a processing unit 132p1 that calculates the gap between the pseudo image and the input image, and a processing unit 132p2 that compares the gap with a threshold, and determines whether the molded product 3 is good or bad based on the input image, the calculated image, and the pseudo image.

[0076] The image generation unit 132a of the first embodiment receives as input a polarized image having four polarization channels acquired by photographing a molded product to be inspected. The image generation unit 132a then generates a pseudo image from the polarized image having the four polarization channels. Even if the molded product in the input image contains an abnormality, the image generation unit 132a generates a pseudo image that has the characteristics of a normal molded product.

[0077] The image generation unit 132a includes a GAN (Generative Adversarial Network) as a learning model. GAN is a learning model that includes a generator and a discriminator. Images of multiple sample molded products without abnormalities (i.e., normal) are provided in advance as training data for the learning model, and machine learning is performed using the images. The multiple sample molded products should preferably include many sample molded products with different states within a normal range. Through such machine learning, the image generation unit 132a generates a pseudo image that has the characteristics of a normal molded product.

[0078] Once the pseudo image is obtained, the determination unit 132 (processing unit 132p1) calculates the difference between the input image and the pseudo image. If the difference is equal to or less than a threshold, the determination unit 132 (processing unit 132p2) determines that the molded product is normal, and if the difference exceeds the threshold, the determination unit 132 (processing unit 132p2) determines that the molded product is abnormal.

[0079] In the process of outputting the determination result (S507), the output unit 133 may output the input image, the pseudo image, and the determination result of normality or abnormality to the display unit 52. In addition, the determination unit 132 may display the gap between the input image and the pseudo image and the locations where the gap is large for each pixel.

[0080] The difference between the input image and the pseudo-image can be expressed by any of a variety of quantities. For example, the following three quantities can be used:

[0081] First, the difference between the input image and the pseudo image can be used as the above-mentioned gap. The difference is the sum of the differences (absolute value or square value of the difference) between the values ​​(brightness values) of the same pixel in the same polarization channel, calculated for all polarization channels and all pixels.

[0082] Second, the distance can be calculated using the value of a certain variable (multiple sets of values) in the latent variable space of the GAN of the image generation unit 132a. As the certain variable in the latent variable space, a variable that reflects the distance between the input image and the pseudo image in the latent variable space can be found and used. When the value of the variable is used, the sum of the distance calculated from the variable and a value obtained by applying a certain weight to the sum of the image differences can be used as a loss function representing the distance.

[0083] Third, the difference in the above gap can be measured by using well-known image features. Classic image features include AKAZE (Accelerated KAZE), cosine similarity, and histograms.

[0084] According to the judgment process of the first embodiment, a pseudo image having the characteristics of a normal product is generated using an image generation unit 132a including a machine-learned learning model (GAN). Therefore, the judgment unit 132 can obtain a pseudo image of the molded product to be inspected that is normal with little load. Then, the judgment unit 132 can accurately judge whether the molded product is normal by calculating the difference between the input image of the molded product to be inspected and the generated pseudo image.

[0085] Furthermore, according to the judgment process of the first embodiment, polarized images having four polarization channels are used as the input image and pseudo image. Molded products produced by the injection molding machine 2 generally transmit light, and their quality is determined by the magnitude of internal stress. Because the polarized images reflect the internal stress of light-transmitting molded products, applying the polarized images as the input image and pseudo image enables highly accurate quality judgment of molded products produced by the injection molding machine 2.

[0086] (Determination unit of the second embodiment) 10, the determination unit 132 includes a processing unit 132p0 that calculates an input image from a polarized image, and an image generation unit 132a that generates a pseudo image from the input image in accordance with a machine-learned learning model. The determination unit 132 also includes a processing unit 132p1 that calculates the gap between the pseudo image and the input image, and a processing unit 132p2 that compares the gap with a threshold, and determines whether the molded product 3 is good or bad based on the input image, the calculated image, and the pseudo image.

[0087] In the second embodiment, the image generating unit 132a receives three input images of the molded product to be inspected: a monochrome image, a linear polarization degree image, and a linear polarization angle image. The image generating unit 132a then generates the monochrome image, the linear polarization degree image, and the linear polarization angle image as pseudo images. Even if the molded product in the input image contains an abnormality, the image generating unit 132a generates a pseudo image that has the characteristics of a normal molded product. Specifically, the three images, the monochrome image, the linear polarization degree image, and the linear polarization angle image, refer to a single image in which the pixel value of the monochrome image, the pixel value of the polarization degree, and the pixel value of the polarization angle are assigned to three channels for each pixel.

[0088] The monochrome image, linear polarization degree image, and linear polarization angle image correspond to calculated images obtained by calculation from a polarization image having four polarization channels. The monochrome image is an image in which the pixel value is the average value of the values ​​of the same pixel in the four polarization channels. The linear polarization degree image is an image in which the value of each pixel corresponds to the degree of linear polarization (DoLP) at each pixel. The degree of linear polarization (DoLP) can be calculated from the values ​​of the four polarization channels. The linear polarization angle image is an image in which the value of each pixel corresponds to the angle of linear polarization (AoLP) at each pixel. The angle of linear polarization (AoLP) can be calculated from the values ​​of the four polarization channels. The calculated images may be generated by the determination unit 132 (processing unit 132p0) or the imaging device 110.

[0089] The image generation unit 132a includes a GAN as a learning model. Images of multiple sample molded products without abnormalities (i.e., normal) are provided in advance as training data for the learning model, and machine learning is performed using these images. The multiple sample molded products should preferably include many sample molded products with different states within a normal range. Through such machine learning, the image generation unit 132a generates a pseudo image having the characteristics of a normal molded product.

[0090] Once the pseudo image is obtained, the determination unit 132 (processing unit 132p1) calculates the difference between the input image and the pseudo image. If the difference is equal to or less than a threshold, the determination unit 132 (processing unit 132p2) determines that the molded product is normal, and if the difference exceeds the threshold, the determination unit 132 (processing unit 132p2) determines that the molded product is abnormal.

[0091] In the process of outputting the determination result (S507), the output unit 133 may output the input image, the pseudo image, and the determination result of normality or abnormality to the display unit 52. In addition, the determination unit 132 may display the gap between the input image and the pseudo image and the locations where the gap is large for each pixel.

[0092] The difference between the input image and the pseudo image can be any quantity that can represent the difference. For example, the three quantities described above can be used. However, of the three quantities described above, the first quantity representing the difference is changed to the sum of the differences between the input image and the pseudo image when the monochrome image component, the linear polarization degree image component, and the linear polarization angle image component are used as the three channel components.

[0093] According to the judgment process of the second embodiment, a pseudo image having the characteristics of a normal product is generated using an image generation unit 132a including a machine-learned learning model (GAN). Therefore, the judgment unit 132 can obtain a pseudo image of the molded product to be inspected that is normal with little load. Then, the judgment unit 132 can accurately judge whether the molded product is normal by calculating the difference between the input image of the molded product to be inspected and the generated pseudo image.

[0094] Furthermore, according to the judgment process of the second embodiment, a set of three images, a monochrome image, a linear polarization degree image, and a linear polarization angle image, calculated from a polarized image having four polarization channels, are used as the input image and pseudo image. Molded products produced by the injection molding machine 2 generally transmit light, and their quality is determined by the amount of internal stress. Because the monochrome image, linear polarization degree image, and linear polarization angle image reflect the internal stress of the light-transmitting molded product, using these images as the input image and pseudo image enables highly accurate quality judgment of the molded products produced by the injection molding machine 2.

[0095] (Determination unit of the third embodiment) 11, the determination unit 132 includes a processing unit 132p0 that calculates an input image from a polarized image, and an image generation unit 132a that generates a pseudo image from the input image in accordance with a machine-learned learning model. The determination unit 132 further includes a processing unit 132p1 that calculates the gap between the pseudo image and the input image, and a processing unit 132p2 that compares the gap with a threshold, and determines whether the molded product 3 is good or bad based on the input image, the calculated image, and the pseudo image.

[0096] The image generating unit 132a of the third embodiment inputs a principal stress-plane-related image of the molded product to be inspected as an input image. Then, the image generating unit 132a generates the stress-related image as a pseudo image. Even if the molded product in the input image contains an abnormality, the image generating unit 132a generates a pseudo image having the characteristics of a normal molded product.

[0097] Principal stress plane-related images are images related to the principal stress plane of the internal stress of a molded product, such as phase contrast images and principal stress difference images. A phase contrast image is an image in which pixel values ​​indicate a phase difference. Here, phase difference refers to the phase difference between polarized light perpendicular to the principal stress plane and polarized light parallel to it. A principal stress difference image is an image in which pixel values ​​indicate a principal stress difference. Principal stress plane-related images (phase contrast images and principal stress difference images) correspond to calculated images obtained by calculation from polarized images having four polarization channels. The generation of the calculated images may be performed by the determination unit 132 (processing unit 132p0) or the imaging device 110.

[0098] As the principal stress plane-related image, an image that further includes the above-mentioned monochrome image as a component of another channel may be applied.

[0099] The image generation unit 132a includes a GAN as a learning model. Images of multiple sample molded products without abnormalities (i.e., normal) are provided in advance as training data for the learning model, and machine learning is performed using these images. The multiple sample molded products should preferably include many sample molded products with different states within a normal range. Through such machine learning, the image generation unit 132a generates a pseudo image having the characteristics of a normal molded product.

[0100] Once the pseudo image is obtained, the determination unit 132 (processing unit 132p1) calculates the difference between the input image and the pseudo image. If the difference is equal to or less than a threshold, the determination unit 132 (processing unit 132p2) determines that the molded product is normal, and if the difference exceeds the threshold, the determination unit 132 (processing unit 132p2) determines that the molded product is abnormal.

[0101] In the process of outputting the determination result (S507), the output unit 133 may output the input image, the pseudo image, and the determination result of normality or abnormality to the display unit 52. In addition, the determination unit 132 may display the gap between the input image and the pseudo image and the locations where the gap is large for each pixel.

[0102] The difference between the input image and the pseudo image can be any quantity that can represent the difference. For example, the three quantities described above can be used. However, the first of the three quantities that represents the difference is changed to the sum of the differences calculated for each pixel of the principal stress plane-related image.

[0103] According to the determination process of the third embodiment, a pseudo image having the characteristics of a normal product is generated using an image generation unit 132a including a machine-learned learning model (GAN). Therefore, the determination unit 132 can obtain a pseudo image of the molded product being inspected that is normal with little load. The determination unit 132 can then accurately determine whether the molded product is normal by calculating the difference between the input image of the molded product being inspected and the generated pseudo image.

[0104] Furthermore, according to the judgment process of the third embodiment, principal stress plane-related images are used as the input image and pseudo image. Molded products manufactured by the injection molding machine 2 are generally light-transmitting, and their quality is determined by the magnitude of internal stress. Since the principal stress plane-related images reflect the internal stress of light-transmitting molded products, applying the principal stress plane-related images as the input image and pseudo image makes it possible to judge the quality of molded products manufactured by the injection molding machine 2 with high accuracy.

[0105] (Determination unit of the fourth embodiment) 12, the determination unit 132 includes a processing unit 132p0 that calculates an input image from a polarized image, and an image generation unit 132a that generates a pseudo image from the input image in accordance with a machine-learned learning model. The determination unit 132 also includes a processing unit 132p1 that calculates the gap between the pseudo image and the input image, and a processing unit 132p2 that compares the gap with a threshold, and determines whether the molded product 3 is good or bad based on the input image, the calculated image, and the pseudo image.

[0106] In the fourth embodiment, the image generating unit 132a receives a relative fringe order image of the molded product to be inspected as an input image. The image generating unit 132a then generates the relative fringe order image as a pseudo image. Even if the molded product in the input image contains an abnormality, the image generating unit 132a generates a pseudo image that has the characteristics of a normal molded product.

[0107] The relative fringe order image is an image of photoelastic fringes obtained by a polarizer, and corresponds to a calculated image obtained by calculation from a polarization image having four polarization channels. The generation of the calculated image may be performed by the determination unit 132 (processing unit 132p0) or the imaging device 110.

[0108] The image generation unit 132a includes a GAN as a learning model. Images of multiple sample molded products without abnormalities (i.e., normal) are provided in advance as training data for the learning model, and machine learning is performed using these images. The multiple sample molded products should preferably include many sample molded products with different states within a normal range. Through such machine learning, the image generation unit 132a generates a pseudo image having the characteristics of a normal molded product.

[0109] Once the pseudo image is obtained, the determination unit 132 (processing unit 132p1) calculates the difference between the input image and the pseudo image. If the difference is equal to or less than a threshold, the determination unit 132 (processing unit 132p2) determines that the molded product is normal, and if the difference exceeds the threshold, the determination unit 132 (processing unit 132p2) determines that the molded product is abnormal.

[0110] In the process of outputting the determination result (S507), the output unit 133 may output the input image, the pseudo image, and the determination result of normality or abnormality to the display unit 52. In addition, the determination unit 132 may display the gap between the input image and the pseudo image and the locations where the gap is large for each pixel.

[0111] The difference between the input image and the pseudo-image can be expressed by any of a variety of quantities, including the three quantities described above. However, the first of the three quantities is changed to the sum of the differences calculated for each pixel in the relative fringe order image.

[0112] According to the judgment process of the fourth embodiment, a pseudo image having characteristics of a normal product is generated using an image generation unit 132a including a machine-learned learning model (GAN). Therefore, the judgment unit 132 can obtain a pseudo image of the molded product to be inspected that is normal with little load. Then, the judgment unit 132 can accurately judge whether the molded product is normal by calculating the difference between the input image of the molded product to be inspected and the generated pseudo image.

[0113] Furthermore, according to the judgment process of the fourth embodiment, a relative fringe order image is used as the input image and pseudo image. Many molded products manufactured by the injection molding machine 2 are generally light-transmitting, and their quality is determined by the magnitude of internal stress. Since the relative fringe order image reflects the internal stress of a molded product that transmits light, applying the relative fringe order image as the input image and pseudo image enables highly accurate quality judgment of molded products manufactured by the injection molding machine 2.

[0114] (Determination unit of the fifth embodiment) 13, the determination unit 132 includes a processing unit 132p0 that calculates a relative fringe order image from the polarized image, and a first image generation unit 132b and a second image generation unit 132c that generate images according to a machine-learned learning model. The determination unit 132 further includes a processing unit 132p1 that calculates the gap between the input image of the second image generation unit 132c and the pseudo image, and a processing unit 132p2 that compares the gap with a threshold, and determines whether the molded product 3 is good or bad based on the input image, the calculated image, and the pseudo image.

[0115] The first image generation unit 132b inputs a relative fringe order image of the molded product to be inspected and outputs a stress distribution image that reflects the stress distribution of the molded product with high accuracy. The first image generation unit 132b includes a GAN as a first learning model. The first learning model is provided with relative fringe order images and stress distribution images of multiple sample molded products as training data in advance, and is machine-learned so that when a relative fringe order image is input, a corresponding stress distribution image is output. The multiple sample molded products preferably include many sample molded products with different conditions ranging from normal to abnormal. Through this machine learning, the first image generation unit 132b can output a stress distribution image that reflects the stress distribution of the molded product to be inspected with high accuracy.

[0116] The second image generating unit 132c receives a stress distribution image of the molded product to be inspected as an input image, and the image generating unit 132a generates a pseudo image of the stress distribution image having the characteristics of a normal molded product, even if the molded product to be inspected contains an abnormality.

[0117] The stress distribution image input to the second image generation unit 132c is an image generated by the first image generation unit 132b based on the relative fringe order image calculated from the polarization images of the four polarization channels. Therefore, the stress distribution image input to the second image generation unit 132c corresponds to a calculated image calculated from the polarization images of the four polarization channels.

[0118] The second image generation unit 132c includes a GAN as a second learning model. Images of multiple sample molded products without abnormalities (i.e., normal) are provided in advance as training data for the second learning model, and machine learning is performed using these images. The multiple sample molded products should preferably include many sample molded products with different states within a normal range. Through such machine learning, the second image generation unit 132c generates a pseudo image having the characteristics of a normal molded product.

[0119] Once the pseudo image is obtained, the determination unit 132 (processing unit 132p1) calculates the difference between the input image and the pseudo image. If the difference is equal to or less than a threshold, the determination unit 132 (processing unit 132p2) determines that the molded product is normal, and if the difference exceeds the threshold, the determination unit 132 (processing unit 132p2) determines that the molded product is abnormal.

[0120] 14 is an image diagram showing an example display of the inspection results. In the determination result output process (S507), the output unit 133 may output the input image pin to the second image generation unit 132c, the pseudo image pout output by the second image generation unit 132c, and the determination result e1 of normal or abnormal to the display unit 52. In addition, the determination unit 132 may display the distance e2 between the input image pin and the pseudo image pout and the areas where the distance is large for each pixel e3.

[0121] The difference between the input image and the pseudo image can be any quantity that can represent the difference. For example, the three quantities mentioned above can be used. However, the first of the three quantities that represents the difference is changed to the sum of the differences calculated for each pixel value of the stress distribution image.

[0122] According to the determination process of the fifth embodiment, a pseudo image having characteristics of a normal product is generated using a second image generation unit 132c including a machine-learned second learning model (GAN). Therefore, the determination unit 132 can obtain a pseudo image of the molded product to be inspected that is normal with little load. The determination unit 132 can then accurately determine whether the molded product is normal by calculating the difference between the input image of the molded product to be inspected and the generated pseudo image.

[0123] Furthermore, according to the judgment process of the fifth embodiment, stress distribution images are used as the input image and pseudo image. The quality of molded products produced by the injection molding machine 2 is determined by the magnitude of the internal stress. Since the stress distribution image reflects the internal stress, applying the stress distribution image as the input image and pseudo image makes it possible to judge the quality of molded products produced by the injection molding machine 2 with high accuracy.

[0124] Machine learning processing Fig. 15 is a flowchart showing an example of the procedure of machine learning processing executed by an inspection device. Fig. 16(A) is a diagram explaining the machine learning processing of the first to fourth embodiments. Fig. 16(B) is a diagram explaining the machine learning processing of the fifth embodiment.

[0125] The image generating unit 132a, the first image generating unit 132b, and the second image generating unit 132c described above need to be given training data in advance to perform machine learning.

[0126] 2, the inspection device 100 has an operation unit 141 that can select an operation mode for performing machine learning (machine learning mode), and a training data input unit 142 that can input training data from outside. Furthermore, the inspection device 100 has a CAE calculation unit 143 that creates training data, and a molding process data input unit 144 that can input molding process data.

[0127] As shown in FIG. 15, the operator can proceed to machine learning processing by selecting a machine learning mode via the operation unit 141 (S601). When proceeding to machine learning processing, the operator inputs training data from the training data input unit 142 (S602). Alternatively, the operator may input molding process data from the molding process data input unit 144 (S603). Then, the CAE calculation unit 143 generates training data based on the molding process data (S604). Then, the determination unit 132 receives the training data and causes the image generation unit 132a (or the first image generation unit 132b and the second image generation unit 132c) to undergo machine learning.

[0128] Machine learning may be performed each time there is a change in the shape, material, or both of the molded product manufactured by the injection molding machine 2. A change in the shape of the molded product means a change in the mold.

[0129] The images that are training data may be polarized images obtained by actually capturing an image of a sample molded article using the imaging device 110, or calculated images obtained by performing calculations (including analysis) on the polarized images. The sample molded article may be a molded article having the same shape and material as the molded article 3 to be inspected.

[0130] Additionally, the images that are training data may be images (virtual calculation images showing stress distribution) of a simulated sample molded product obtained by calculation based on the characteristics of the simulated sample molded product (virtual molded product) that are obtained by performing CAE (Computer Aided Engineering) calculations based on molding process data of the injection molding machine 2. The molding process data is data that can simulate the molding process, and includes at least data on the cavity shape of the mold.

[0131] In the example of Figure 2, the inspection device 100 is shown to be equipped with a CAE calculation unit 143 that performs the above-mentioned CAE calculations and to be configured to create images of training data within the inspection device 100, but the CAE calculation unit 143 may also be configured as a computer separate from the inspection device 100.

[0132] In the first embodiment, the training data used are polarization images of multiple sample molded articles, each containing four polarization channels. The multiple sample molded articles are normal products, and their states should differ within a normal range.

[0133] In the second embodiment, a set of monochrome images, linear polarization degree images, and linear polarization angle images of multiple sample molded articles is used as training data. The multiple sample molded articles are normal products, and it is preferable that their states differ within a normal range.

[0134] In the third embodiment, images related to the principal stress planes of multiple sample molded articles (phase contrast images, principal stress difference images, etc.) are used as training data. The multiple sample molded articles are normal products, and it is preferable that their states differ within a normal range.

[0135] In the fourth embodiment, relative fringe order images of a plurality of sample molded articles are used as training data. The plurality of sample molded articles are normal products, and it is preferable that the states of the sample molded articles differ within a normal range.

[0136] In the fifth embodiment, the training data for the first image generation unit 132b is a combination of a relative fringe order image and a stress distribution image for each of a plurality of sample molded products. The first image generation unit 132b is trained by machine learning so that it can input a relative fringe order image and output a corresponding stress distribution image. The plurality of sample molded products may include both normal and abnormal products and may be in different states.

[0137] Stress distribution images of a plurality of sample molded articles are applied as training data for the second image generating unit 132c. The plurality of sample molded articles are normal articles, and it is preferable that the states of the sample molded articles are different within a normal range.

[0138] The above-mentioned machine learning enables the determination unit 132 to perform the above-mentioned determination process.

[0139] As described above, in the inspection device 100 and injection molding system 1 of this embodiment, the imaging device 110 acquires a polarized image of the molded product 3 to be inspected. Furthermore, the image generation unit 132a or the second image generation unit 132c of the determination unit 132 inputs the polarized image or a calculated image obtained by calculation from the polarized image as an input image, and outputs a pseudo image having characteristics of a normal product. The determination unit 132 then determines whether the molded product is good or bad based on the input image and the pseudo image. Therefore, the determination unit 132 can perform a high-accuracy pass / fail determination with a small load.

[0140] Furthermore, the inspection device 100 has a training data input unit 142 for inputting training data, and the training data includes at least a polarized image of a sample molded product that is a normal product, a computed image obtained by computing the polarized image, a virtual polarized image of a virtual molded product calculated by simulating the process using CAE calculations based on molding process data, and a computed image obtained by computing the virtual polarized image.The image generation unit 132a (or the second image generation unit 132c) then performs machine learning using the training data.This machine learning allows the generation of the pseudo image described above to be realized.

[0141] Furthermore, the polarized image or virtual polarized image includes four or more polarization channels with different polarization angles. Furthermore, the calculated image includes one or more of a monochrome image, a combination of a linear polarization degree image and a linear polarization angle image, a principal stress plane-related image, a relative fringe order image, and a stress distribution image. Using such images, highly accurate pass / fail judgment can be performed on molded products molded by the injection molding machine 2 that transmit light.

[0142] The embodiments of the present invention have been described above. However, the present invention is not limited to the above embodiments. The details shown in the embodiments can be modified as appropriate without departing from the spirit of the invention. [Explanation of symbols]

[0143] 1. Injection molding system 2 Injection molding machine 3 Molded products 10 Injection device 20 screws 40 Control device 60 Drive Unit 61 Metering motor 71 Injection motor 100 Inspection equipment 110 Imaging device 111 Light source 112 Linear polarizer 113 Wave plate 114 Polarized Camera 120 Processing equipment 131 Receiving unit 132 Judgment section 132a Image generation unit 132b First image generation unit 132c Second image generation unit 133 Output section 215,315 Rotation means 141 Operation section 142 Training data input section 143 CAE calculation department 144 Molding process data input section pin input image pout pseudo image e1 Judgment result

Claims

1. an imaging device for acquiring polarization images of a molded article in a plurality of polarization channels; a judgment unit for judging whether the molded product is good or bad; Equipped with The determination unit an image generation unit that generates a pseudo image from an input image in accordance with a machine-learned learning model; inputting the polarized image or a calculated image obtained by calculation from the polarized image into the image generation unit as the input image, and determining whether the molded product is good or bad based on the input image and the pseudo image generated by the image generation unit; the image generation unit generates pseudo images of polarization channels having the same number of channels from the polarization images or the calculated images of the plurality of polarization channels; the determining unit compares the polarization image or the calculation image with the pseudo image for the same polarization channel and determines whether the image is good or bad based on the result of the comparison. Inspection equipment.

2. a training data input unit for inputting machine learning training data; the training data includes at least one of a polarized image of a normal molded product, a calculated image obtained by calculation from the polarized image of the normal molded product, a virtual polarized image of a virtual molded product calculated by simulation based on molding process data that can simulate a normal molding process, and a calculated image calculated from the virtual polarized image; The learning model is machine-learned using the input training data. The inspection device according to claim 1.

3. the polarized image or the virtual polarized image includes four or more polarized channels with different polarization angles; The inspection device according to claim 2.

4. The calculated image includes one or more of a monochrome image, a combination of a linear polarization degree image and a linear polarization angle image, a principal stress plane related image, a relative fringe order image, and a stress distribution image. The inspection device according to claim 2.

5. An injection molding system comprising: an injection molding machine; and the inspection device according to any one of claims 1 to 4, which inspects a molded product molded by the injection molding machine.

6. The imaging device acquires polarization images of the molded product from multiple polarization channels, inputting the polarized image or a calculated image obtained by calculation from the polarized image to an image generating unit as an input image; determining whether the molded product is good or bad based on the pseudo image of the input image generated by the image generation unit and the input image; the image generation unit generates pseudo images of polarization channels having the same number of channels from the polarization images or the calculated images of the plurality of polarization channels; comparing the polarized image or the calculated image with the pseudo image between the same polarization channels, and determining whether the image is good or bad based on the result of the comparison; Testing method.

Images