Filament abnormality detection device and abnormality detection program

Abstract

A technique for detecting a thread path abnormality of a filament with respect to a guide is provided. The filament abnormality detection device detects an abnormality in a spinning drawing device that draws a plurality of filaments spun from a spinning device. The spinning drawing device includes a guide for guiding a plurality of filaments conveyed from an upstream side to a downstream side so as to approach each other. The filament abnormality detection device is provided with a control device. The control device executes: an acquisition process of acquiring, from a camera configured to include at least a plurality of filaments passing through a guide within an imaging range, an estimation filament image in which the plurality of filaments are reflected; a determination process for determining whether or not a thread path abnormality has occurred on the basis of a machine-learned estimation model for detecting a thread path abnormality in which at least one of the plurality of filaments is separated from the guide, and the filament image for estimation; and an output process for outputting a determination result in the determination process.

Description

Technical Field

[0001] The present disclosure relates to a filament abnormality detection device and an abnormality detection program. Background Art

[0002] Japanese Unexamined Patent Application Publication No. 2023-128942 (Patent Document 1) discloses a "detection system capable of highly accurately detecting the filament swing of a filamentous body in a spinning process during filament manufacturing".

[0003] This detection system causes a camera unit to photograph a filamentous body extruded from a spinning nozzle, and acquires a plurality of input images in which the filamentous body is reflected. After that, this detection system calculates the degree of change in the gray value for pixels at the same position among the plurality of input images, and detects the filament swing of the filamentous body based on this degree of change.

[0004] Prior Art Documents

[0005] Patent Documents

[0006] Patent Document 1: Japanese Unexamined Patent Application Publication No. 2023-128942

[0007] A spinning and drawing device manufactures a filament by drawing a plurality of filaments spun from a spinning device and twisting the plurality of filaments. The plurality of filaments are brought closer to each other by a guide member during the conveyance in the spinning and drawing device. At this time, at least one of the plurality of filaments sometimes detaches from the guide member. Regarding whether such a filament path abnormality has occurred, in most cases, it is confirmed visually by an operator. However, the filaments are very thin, and visual confirmation is very difficult.

[0008] In view of the above problems, a technique for detecting a filament path abnormality of a filament with respect to a guide member is desired.

[0009] It should be noted that the purpose of the detection system disclosed in Patent Document 1 is to detect the filament swing of a filamentous body, and its purpose is not to detect a filament path abnormality of a filament with respect to a guide member. Summary of the Invention

[0010] In an example of the present disclosure, a filament abnormality detection device is provided that can detect an abnormality in a spinning and drawing device that draws a plurality of filaments spun from a spinning device. The spinning and drawing device includes a guide member that guides the plurality of filaments conveyed from the upstream side to the downstream side in a manner of approaching each other. The filament abnormality detection device includes a control device. The control device performs the following processes: an acquisition process of acquiring a filament image for estimation that reflects the plurality of filaments from a camera configured to include at least the plurality of filaments passing through the guide member within a shooting range; a determination process of determining whether the filament path abnormality has occurred based on a estimation model that has undergone machine learning to detect an abnormality in which at least one of the plurality of filaments detaches from the guide member and the filament image for estimation; and an output process of outputting a determination result in the determination process.

[0011] In the above filament abnormality detection device, an estimation model that has undergone machine learning to detect an abnormality in which a filament detaches from a guide member is used. The filament abnormality detection device can detect a filament path abnormality by inputting the filament image for estimation into the estimation model.

[0012] In an example of the present disclosure, the estimation model is an autoencoder that has undergone machine learning to restore a normal filament image that has not undergone the filament path abnormality after compression. In the determination process, based on the similarity between the filament image for estimation and the restored filament image obtained by inputting the filament image for estimation into the autoencoder, it is determined whether the filament path abnormality has occurred.

[0013] In the above filament abnormality detection device, an autoencoder generated based on normal filament images is used. That is, during machine learning, abnormal filament images are not required.

[0014] In an example of the present disclosure, in the determination process, based on the comparison result between the similarity and a predetermined threshold value, it is determined whether the filament path abnormality has occurred.

[0015] Thus, the filament abnormality detection device can detect a filament path abnormality based on the threshold value.

[0016] In an example of the present disclosure, the estimation model is generated by performing machine learning on a plurality of learning data. In each of the plurality of learning data, a label indicating whether the filament path abnormality has occurred is associated with a learning filament image that reflects a plurality of filaments passing through the guide member. In the determination process, based on the output result obtained from the estimation model by inputting the filament image for estimation into the estimation model, it is determined whether the filament path abnormality has occurred.

[0017] In the above filament abnormality detection device, a presumption model learned using abnormal filament images in which a filament path abnormality has occurred is used. Thereby, the detection accuracy of the filament path abnormality is improved.

[0018] In one example of the present disclosure, the above determination process includes: a process of performing preprocessing on the above-mentioned filament image for presumption; and a process of inputting the preprocessed filament image for presumption into the above-mentioned presumption model. The above preprocessing includes a process of extracting edges from the above-mentioned filament image for presumption.

[0019] Thereby, the filament abnormality detection device can capture the filaments reflected in the filament image for presumption more accurately. As a result, the detection accuracy of the filament path abnormality is improved.

[0020] In one example of the present disclosure, the above spinning and drawing device further includes a light source configured to include at least the above-mentioned multiple filaments passing through a position upstream of the above-mentioned guide member within the illumination range.

[0021] Thereby, the brightness difference between the filaments and the background becomes clear, and the filament abnormality detection device can capture the filaments reflected in the filament image for presumption more accurately. As a result, the detection accuracy of the filament path abnormality is improved.

[0022] In one example of the present disclosure, the above-mentioned guide member includes: a main body having a surface in contact with the above-mentioned multiple filaments conveyed from the above-mentioned spinning device in the direction of gravity; a discharge port formed in the above-mentioned surface for discharging an oil agent; and two filament guiding members provided on the above-mentioned surface so as to be located on both sides in the horizontal direction of the above-mentioned discharge port, and the two filament guiding members are used to guide the above-mentioned multiple filaments conveyed in the above-mentioned direction of gravity toward the central side in the above-mentioned horizontal direction of the above-mentioned discharge port. The above-mentioned camera is configured to include at least the above-mentioned multiple filaments passing through a position upstream of the above-mentioned surface within the shooting range of the camera. The above-mentioned light source is configured to include at least the above-mentioned multiple filaments passing through a position upstream of the above-mentioned surface within the illumination range of the light source.

[0023] Thereby, the filament abnormality detection device can detect a filament path abnormality in a guide member having an oil supply function. In addition, the shooting range of the camera includes a position upstream of the above-mentioned surface, and the illumination range of the light source includes a position upstream of the above-mentioned surface. Thereby, the filament abnormality detection device can more reliably capture the filament path abnormality of the filaments. As a result, the detection accuracy of the filament path abnormality is improved.

[0024] In one example of the present disclosure, the optical axis of the above-mentioned light source is inclined with respect to the optical axis of the above-mentioned camera.

[0025] Thereby, the amount of light reflected toward the camera can be reduced. As a result, the generation of whitening in the filament image can be prevented.

[0026] In one example of the present disclosure, the above light source is disposed at a position closer to the front side than the above guide member in the direction of observing the above guide member from the above camera.

[0027] As a result, the amount of light directly incident from the light source into the camera is reduced. As a result, the generation of whitening in the filament image can be prevented.

[0028] In one example of the present disclosure, the above spinning and drawing device further includes an anti-reflection member for preventing reflection of light irradiated from the above light source. The above anti-reflection member is disposed at a position farther from the above guide member in the direction of observing the above guide member from the above camera.

[0029] As a result, the amount of light reflected toward the camera can be further reduced. As a result, the generation of whitening in the filament image can be more reliably prevented.

[0030] In one example of the present disclosure, the above anti-reflection member is black.

[0031] As a result, the amount of light reflected toward the camera can be further reduced. As a result, the generation of whitening in the filament image can be more reliably prevented.

[0032] In one example of the present disclosure, the above light source is a condenser illumination. The illumination range of the above condenser illumination intersects the optical axis of the above camera.

[0033] As a result, light irradiation to objects other than the filaments can be suppressed. Therefore, the amount of light reflected toward the camera can be further reduced. As a result, the generation of whitening in the filament image can be more reliably prevented.

[0034] In other examples of the present disclosure, there is provided an abnormality detection program executed by a filament information processing device capable of communicating with a spinning and drawing device that draws a plurality of filaments spun from a spinning device. The above spinning and drawing device includes a guide member for guiding the above plurality of filaments closer to each other. The above abnormality detection program causes the filament information processing device to execute the following steps: an acquisition step of acquiring a presumptive filament image that reflects the above plurality of filaments from a camera configured to include at least the above plurality of filaments passing through the above guide member within a shooting range; an input step of inputting the above presumptive filament image to a presumption model that has been machine-learned to presume a filament path abnormality in which at least one of the above plurality of filaments detaches from the above guide member; a determination step of determining whether the above filament path abnormality has occurred based on the output result of the above presumption model; and an output step of outputting the determination result in the above determination step.

[0035] In the above abnormal detection program, a presumption model that has undergone machine learning to detect a filament path abnormality in which a filament detaches from a guide member is used. The above abnormal detection program can detect a filament path abnormality by inputting a filament image for presumption into the presumption model.

[0036] In an example of the present disclosure, the above presumption model is an autoencoder that has undergone machine learning to restore a normal filament image after compressing the normal filament image in which the above filament path abnormality has not occurred. In the above determination step, based on the similarity between the filament image for presumption and the restored filament image obtained by inputting the filament image for presumption into the autoencoder, it is determined whether the above filament path abnormality has occurred.

[0037] In the above abnormal detection program, an autoencoder generated based on a normal filament image is used. That is, during machine learning, abnormal filament images are not required.

[0038] In an example of the present disclosure, the above determination step includes: a step of performing preprocessing on the filament image for presumption; and a step of inputting the preprocessed filament image for presumption into the presumption model. The above preprocessing includes a process of extracting edges from the filament image for presumption.

[0039] As a result, the filament abnormality detection device can capture the filament reflected in the filament image for presumption more accurately. As a result, the detection accuracy of the filament path abnormality is improved.

[0040] The above and other objects, features, aspects, and advantages of the present invention will become apparent from the following detailed description of the present invention understood in association with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 is a schematic diagram showing an example of the device configuration of a spinning and drawing device.

[0042] Figure 2 is a view showing an oil supply guide member from the front.

[0043] Figure 3 is along Figure 2 a cross-sectional view taken along line III-III shown.

[0044] Figure 4 is a view showing a filament path abnormality of a filament.

[0045] Figure 5 is a view showing the main configuration for realizing the detection function of a filament path abnormality.

[0046] Figure 6 is a view showing an example of the hardware configuration of an information processing device.

[0047] Figure 7 This is a diagram showing an example of a learning dataset.

[0048] Figure 8 This is a diagram showing an example of the functional configuration of an information processing device.

[0049] Figure 9 This is a diagram for explaining an example of preprocessing performed by a preprocessing unit.

[0050] Figure 10 This is a diagram for explaining another example of preprocessing performed by a preprocessing unit.

[0051] Figure 11 This is a diagram for explaining an example of learning processing performed by a learning unit.

[0052] Figure 12 This is a diagram for explaining an example of judgment processing performed by a judgment unit.

[0053] Figure 13 This is a diagram for explaining an example of the configuration mode of a camera and a light source.

[0054] Figure 14 This is a diagram for explaining another example of the configuration mode of a camera and a light source.

[0055] Figure 15 This is a diagram for explaining yet another example of the configuration mode of a camera and a light source.

[0056] Figure 16 This is a flowchart showing the process of learning processing.

[0057] Figure 17 This is a flowchart showing the process of anomaly detection processing.

[0058] Figure 18 This is a diagram showing an example of the device configuration of an information processing system.

[0059] Figure 19 This is a diagram showing a learning dataset of a modified example.

[0060] Figure 20 This is a diagram for explaining an example of learning processing of a modified example.

[0061] Figure 21 This is a diagram for explaining an example of judgment processing of a modified example.

[0062] Explanation of reference numerals:

[0063] 1: Spinning and drawing device;

[0064] 30: Camera;

[0065] 40: Light source;

[0066] 60: Anti-reflection component;

[0067] 100: Information processing device;

[0068] 101: Control device;

[0069] 123: Learning data;

[0070] 124: Estimation model;

[0071] 124A: Autoencoder;

[0072] 124B: Estimation model. Detailed implementation manners

[0073] Hereinafter, each embodiment of the present invention will be described with reference to the drawings. In the following description, the same reference numerals are assigned to the same components and elements. Their names and functions are also the same. Therefore, detailed descriptions thereof will not be repeated. In addition, the embodiments and each modification described below can be selectively combined as appropriate.

[0074] <A. Spinning and drawing device 1>

[0075] First, refer to Figure 1 The spinning and drawing device 1 as a manufacturing device for drawn filaments will be described. Figure 1 It is a schematic diagram showing an example of the device configuration of the spinning and drawing device 1.

[0076] As Figure 1 shown, the spinning and drawing device 1 respectively draws a synthetic fiber filament Y composed of a plurality of filaments F spun from a spinning device 2, and respectively winds them around a plurality of bobbins B to form a plurality of packages P. It should be noted that hereinafter, Figure 1 the up-down direction, front-back direction, and left-right direction shown will be defined as the up-down direction, front-back direction, and left-right direction of the spinning and drawing device 1 respectively for description.

[0077] In addition, hereinafter, the direction in which a plurality of filaments F are conveyed in the spinning and drawing device 1 will be defined as the downstream side, and the direction opposite to the direction in which a plurality of filaments F are conveyed in the spinning and drawing device 1 will be defined as the upstream side for description.

[0078] The spinning and drawing device 1 includes a cooling unit 3, an oil supply unit 4, a stretching unit 5, a drawing roller 6, a drawing roller 7, a traversing device 8, a winding device 9, etc. First, in the spinning device 2, the polymer supplied from a polymer supply device (not shown) composed of a gear pump or the like is supplied from the left-right direction ( Figure 1A plurality of spinnerets 2a arranged in the paper surface depth direction (in the downward direction) extrude downward, and each group of filaments F is spun in a state arranged in the left-right direction.

[0079] Then, each group of filaments F is sent to the cooling section 3 and the oil supply section 4. Each group of filaments F is gathered into a single thread Y in the oil supply section 4. Then, multiple threads Y travel in a state arranged in the left-right direction in a thread path along the stretching section 5, the take-up roller 6, the traversing device 8, and the take-up roller 7. And, on the basis of being distributed in the front-rear direction from the take-up roller 7, multiple threads Y are respectively wound around a plurality of bobbins B in the winding device 9.

[0080] The cooling section 3 has a plurality of cylindrical cooling cylinders 10, and each cooling cylinder 10 is respectively arranged below a plurality of spinnerets 2a provided in the spinning device 2. A plurality of filaments F spun from the spinnerets 2a of the spinning device 2 travel downward from above along the axial direction of each cooling cylinder 10 in the internal space 10a of each cooling cylinder 10. A rectifying section 10b is provided around the internal space 10a, and cooling air supplied from a compressed air supply device (not shown) flows into the internal space 10a while being rectified by the rectifying section 10b. The rectifying section 10b mainly rectifies the cooling air flowing into the internal space 10a so that the flow rate of the cooling air in the circumferential direction of the cooling cylinder 10 is substantially uniform.

[0081] The oil supply section 4 has a plurality of oil supply guides 11 respectively arranged below each cooling cylinder 10. The oil supply guides 11 gather a plurality of filaments F spun from the spinnerets 2a into a single thread Y and impart an oil agent to the thread Y (a plurality of filaments F). The oil supply guides 11 will be described in detail later.

[0082] In addition, a camera 30 is provided inside the spinning and drawing device 1. The camera 30 is arranged so that the shooting range 30R includes at least a plurality of filaments F passing through a position upstream of the oil supply guide 11. That is, the camera 30 can be arranged to include the oil supply guide 11 in the shooting range 30R, or can be arranged not to include the oil supply guide 11 in the shooting range 30R.

[0083] In addition, a light source 40 is provided inside the spinning and drawing device 1. The light source 40 is arranged so that the illumination range 40R includes at least a plurality of filaments F passing through a position upstream of the oil supply guide 11. That is, the light source 40 can be arranged to include the oil supply guide 11 in the illumination range 40R, or can be arranged not to include the oil supply guide 11 in the illumination range 40R. The illumination range 40R overlaps with the shooting range 30R at least in part.

[0084] The stretching section 5 is arranged below the oil supply section 4. The stretching section 5 has a heat-insulating box 12 and a plurality of heating rollers (not shown) housed in the heat-insulating box 12. The stretching section 5 heats and stretches multiple threads Y respectively using the plurality of heating rollers.

[0085] The multiple filaments Y stretched by the stretching part 5 are conveyed by the haul-off rollers 6 and 7 to the winding device 9. The traversing device 8 is arranged between the haul-off roller 6 and the haul-off roller 7 to wind the multiple filaments F constituting one filament Y to give traversing.

[0086] The winding device 9 includes a machine base 13, a turntable 14, two bobbin holders 15, a support frame 16, a contact roller 17, a traversing device 18, etc. The winding device 9 winds the multiple filaments Y sent from the haul-off roller 7 onto multiple bobbins B simultaneously by rotating the bobbin holders 15 to form multiple packages P.

[0087] The turntable 14 is a disk-shaped component and is installed on the machine base 13. The turntable 14 is rotationally driven by a motor (not shown). The two bobbin holders 15 are cantilever-supported on the turntable 14 in a posture extending in the front-rear direction. A plurality of cylindrical bobbins B are installed on each bobbin holder 15 in a state arranged along its axial direction. By rotating the turntable 14, the two bobbin holders 15 can be switched between the upper winding position and the lower retracted position.

[0088] The support frame 16 is a long frame-shaped component extending in the front-rear direction. The support frame 16 is fixed to the machine base 13. A long roller support member 19 extending in the front-rear direction is installed at the lower part of the support frame 16 in a manner capable of moving up and down relative to the support frame 16. The contact roller 17 extending along the axial direction of the bobbin holder 15 is rotatably supported by the roller support member 19. The contact roller 17 contacts the package P being formed and applies a prescribed contact pressure to the package P, thereby adjusting the shape of the package P.

[0089] The traversing device 18 has a plurality of traversing guides 18a arranged in the front-rear direction. The plurality of traversing guides 18a are driven by a motor (not shown) and reciprocate in the front-rear direction respectively. The traversing guides 18a reciprocate in a state where the filament Y is hung, whereby the filament Y traverses back and forth with the fulcrum guide 18b as the center and is wound around the corresponding bobbin B.

[0090] <B. Oil supply guide 11>

[0091] Next, with reference to Figure 2 and Figure 3 , the Figure 1 shown oil supply guide 11 will be described. Figure 2 is a view showing the oil supply guide 11 from the front. Figure 3 is a cross-sectional view along the Figure 2 shown III-III line.

[0092] As described above, the oil supply guide 11 applies an oil agent to the thread Y composed of a plurality of filaments F spun from the spinning device 2. The oil supply guide 11 is formed of a ceramic material such as alumina or zirconia, and as shown in Figure 2 , Figure 3 , it has a guide body 20. The front surface 21 of the guide body 20 extends in the vertical direction. And, the thread Y (a plurality of filaments F) traveling from the cooling unit 3 in the direction from above to below (i.e., the direction of gravity) contacts the surface 21. The above-described camera 30 is arranged such that its shooting range 30R includes a plurality of filaments F passing through a position upstream of the surface 21. In addition, the above-described light source 40 is arranged such that its illumination range 40R includes a plurality of filaments F passing through a position upstream of the surface 21.

[0093] In addition, the guide body 20 has an oil agent flow path 22. The oil agent flow path 22 is formed inside the oil supply guide 11 and extends in the front-rear direction. The front end of the oil agent flow path 22 becomes a discharge port 25 formed on the surface 21, and the oil agent is discharged through the discharge port 25 to apply the oil agent to the thread Y (a plurality of filaments F). The concentration of the oil agent discharged from the discharge port 25 is, for example, about 85%. The concentration of the oil agent refers to the total concentration of effective components including oil components, additives, etc. other than water.

[0094] Here, the surface 21 of the guide body 20 has an upper curved surface 26 above the upper end 25a of the discharge port 25 and a lower curved surface 27 below the lower end 25b of the discharge port 25. Both the curved surfaces 26 and 27 are curved so as to protrude outward from the guide body 20.

[0095] In addition, when viewed from the left-right direction ( Figure 3 cross-section), the upper end 25a of the discharge port 25 and the upper portion 20a of the guide body 20 above the upper end 25a of the discharge port 25 do not overlap with the tangent line L1 of the lower curved surface 27 at the position of the lower end 25b of the discharge port 25. Moreover, when viewed from the left-right direction, the upper end 25a of the discharge port 25 and the upper portion 20a of the guide body 20 also do not overlap with the straight line L2 obtained by rotating the tangent line L1 10° in the clockwise direction (the direction approaching the upper end 25a of the discharge port 25) around the lower end 25b of the discharge port 25. Figure 3 The length K between the upper end 25a and the lower end 25b of the discharge port 25 is, for example, about 0.1 [mm] in the direction orthogonal to the tangent line L1, whereby the upper end 25a and the lower end 25b of the discharge port 25 have the above-described positional relationship.

[0096] In addition, the oil supply guide 11 is arranged such that when viewed from the left-right direction, the tangent line L1 is substantially parallel to the traveling direction of the thread Y (filament F) sent from the cooling unit 3.

[0097] In addition, the oil supply guide 11 is arranged such that when viewed from the left-right direction, the tangent line L1 is substantially parallel to the traveling direction of the thread Y (filament F) sent from the cooling unit 3.

[0098] In addition, the oil supply guide member 11 is configured to guide a plurality of filaments F conveyed from the upstream side to the downstream side in a manner that they approach each other. More specifically, two wire guide members 23 are arranged on the surface 21 of the guide member main body 20. The two wire guide members 23 are respectively arranged in a portion of the surface 21 on the right side of the discharge port 25 and a portion on the left side of the discharge port 25. That is, the two wire guide members 23 are arranged on the surface 21 so as to be located on both sides in the left-right direction of the discharge port 25. In addition, the two wire guide members 23 extend obliquely with respect to the up-down direction in such a manner that they approach the central portion in the left-right direction of the discharge port 25 more as they go from above to below. Thus, the interval in the left-right direction between the two wire guide members 23 becomes smaller as they go from above to below. And, during the plurality of filaments F sent from the cooling unit 3 pass through the oil supply guide member 11, they are guided by the two wire guide members 23 toward the central side in the left-right direction close to the discharge port 25, and thus gradually converge and gather to become a single wire Y.

[0099] <C. Summary>

[0100] As described above, the plurality of filaments F are pulled closer to each other by the oil supply guide member 11 during the process of being conveyed inside the spinning and drawing device 1. At this time, at least one of the plurality of filaments F sometimes detaches from the oil supply guide member 11. Hereinafter, the detachment of at least one of the plurality of filaments F from the normal filament path on the oil supply guide member 11 is also referred to as "filament path abnormality".

[0101] Figure 4 It is a diagram showing a specific example of the filament path abnormality of the filament F. In Figure 4 In the example of (A), the filament FA detaches from the wire guide member 23 of the oil supply guide member 11. In Figure 4 In the example of (B), the filament FB breaks for some reason and detaches from the wire guide member 23 of the oil supply guide member 11. As other examples, there are also filament path abnormalities such as a filament detaching from the oil supply guide member that it should originally pass through and erroneously entering an adjacent oil supply guide member.

[0102] If the above-mentioned filament path abnormality occurs, the quality of the produced wire Y will deteriorate. Whether a filament path abnormality has occurred is mostly confirmed by the operator's visual inspection. However, the filament F is very thin, and it is very difficult to confirm by visual inspection. Therefore, the inventor has studied a function for automatically detecting the filament path abnormality of the filament F.

[0103] Refer to Figure 5 to explain the outline of the detection function of the filament path abnormality. Figure 5 It is a diagram showing the main configuration of the filament abnormality detection device 50 of the embodiment.

[0104] As shown inFigure 5 As shown, the filament abnormality detection device 50 includes the above-described camera 30 and the information processing device 100. The camera 30 of the filament abnormality detection device 50 is arranged inside the spinning and drawing device 1 in such a manner that at least multiple filaments F passing through the oil supply guide 11 of the spinning and drawing device 1 are included within the shooting range. The detection function for filament path abnormalities is installed, for example, in the information processing device 100. The information processing device 100 is configured to be able to communicate with the spinning and drawing device 1. The information processing device 100 can be arranged inside the spinning and drawing device 1 or outside the spinning and drawing device 1.

[0105] The information processing device 100 is, for example, the control unit of the spinning and drawing device 1. This control unit controls various drive devices (such as the winding device 9, etc.) provided in the spinning and drawing device 1. As another example, the information processing device 100 can also be a server configured to be able to communicate with the spinning and drawing device 1.

[0106] The information processing device 100 is provided with a control device 101. The control device 101 is composed of, for example, at least one integrated circuit. The integrated circuit can be composed of, for example, at least one CPU (Central Processing Unit), at least one GPU (Graphics Processing Unit), at least one ASIC (Application Specific Integrated Circuit), at least one FPGA (Field Programmable Gate Array), or a combination thereof, etc.

[0107] First, the control device 101 obtains an image (hereinafter also referred to as "filament image") in which multiple filaments F are reflected by sending a shooting instruction to the camera 30. In addition, the control device 101 obtains a prediction model 124. The prediction model 124 has been pre-trained by machine learning so as to be able to detect filament path abnormalities. The learning process for generating the prediction model 124 will be described later. Then, the control device 101 determines whether a filament path abnormality has occurred based on the prediction model 124 and the prediction filament image obtained from the camera 30, and outputs the determination result.

[0108] As described above, the information processing device 100 can detect filament path abnormalities by using the prediction model 124 that has been trained by machine learning to detect filament path abnormalities where the filaments F are detached from the oil supply guide 11.

[0109] <D. Hardware Configuration of Information Processing Device 100>

[0110] Next, referring to Figure 6 , to Figure 5The hardware configuration of the information processing apparatus 100 shown will be described. Figure 6 This is a diagram showing an example of the hardware configuration of the information processing apparatus 100.

[0111] The information processing apparatus 100 includes the control device 101, a ROM (Read Only Memory) 102, a RAM (Random Access Memory) 103, a communication interface 104, a display interface 105, an input interface 107, and an auxiliary storage device 120. These components are connected to a bus 115.

[0112] The control device 101 controls the operation of the information processing apparatus 100 by executing various programs such as a learning program 126 and an anomaly detection program 128. Based on receiving an execution command for various programs, the control device 101 reads out the program to be executed from the auxiliary storage device 120 or the ROM 102 to the RAM 103. The RAM 103 functions as a working memory and temporarily stores various data required for executing the program.

[0113] The communication interface 104 is an interface for the information processing apparatus 100 to communicate with external devices. The information processing apparatus 100 exchanges data with external devices via the communication interface 104. Such external devices include, for example, the above-described camera 30 and the above-described light source 40.

[0114] A display 106 is connected to the display interface 105. The display interface 105 sends an image signal for displaying an image to the display 106 in accordance with instructions from the control device 101 and the like. The display 106 is, for example, a liquid crystal display, an organic EL (ElectroLuminescence) display, or other displays. It should be noted that the display 106 may be integrally formed with the information processing apparatus 100 or may be formed separately from the information processing apparatus 100.

[0115] An input device 108 is connected to the input interface 107. The input device 108 is, for example, a mouse, a keyboard, a touch panel, or other devices capable of accepting user operations. It should be noted that the input device 108 may be integrally formed with the information processing apparatus 100 or may be formed separately from the information processing apparatus 100.

[0116] The auxiliary storage device 120 is, for example, a storage medium such as a hard disk, a flash memory, and an SSD (Solid State Drive). The auxiliary storage device 120 stores, for example, the learning dataset 122, the above-mentioned estimation model 124, the learning program 126, and the anomaly detection program 128. Their storage locations are not limited to the auxiliary storage device 120 and can also be stored in the storage area of the control device 101 (for example, a cache memory, etc.), the ROM 102, the RAM 103, an external device (for example, a server), etc.

[0117] The learning program 126 is a program for generating the estimation model 124 using the learning dataset 122. The learning program 126 may not be a single program but may be provided as a part incorporated into an arbitrary program. In this case, the learning process performed by the learning program 126 is realized in cooperation with an arbitrary program. Even such a program that does not include some modules does not deviate from the gist of the learning program 126 according to the present embodiment. Moreover, part or all of the functions provided by the learning program 126 can also be realized by dedicated hardware. In addition, the information processing device 100 may be configured in such a way that at least one server executes a part of the processing of the learning program 126, that is, in the form of a so-called cloud service.

[0118] The anomaly detection program 128 is a program for detecting a filament path anomaly of the filament F using the estimation model 124. The anomaly detection program 128 may not be a single program but may be embedded as a part of an arbitrary program. In this case, the estimation process performed by the anomaly detection program 128 is realized in cooperation with an arbitrary program. Even such a program that does not include some modules does not deviate from the gist of the anomaly detection program 128 of the present embodiment. And part or all of the functions provided by the anomaly detection program 128 can also be realized by dedicated hardware. And the information processing device 100 may be configured in such a way that at least one server executes a part of the processing of the anomaly detection program 128, that is, in the form of a so-called cloud service.

[0119] <E. Learning Dataset 122>

[0120] Next, with reference to Figure 7 , for generating Figure 6 the learning dataset 122 used when the shown estimation model 124 is described. Figure 7 is a diagram showing an example of the learning dataset 122.

[0121] The learning dataset 122 includes a plurality of learning data 123. The number of learning data 123 included in the learning dataset 122 is arbitrary. As an example, the number of learning data 123 is several tens to several hundreds of thousands.

[0122] In each piece of learning data 123, an image name is associated with the filament image. The image name is an identifier for uniquely identifying the filament image.

[0123] In the present embodiment, the learning data set 122 is composed of normal filament images in which the filament F has not detached from the oil supply guide member 11. The normal filament image shows the state in which the filament F travels inside the two wire guide members 23 (see Figure 2 ).

[0124] The learning data set 122 may include filament images acquired from the camera 30 in the spinning and drawing device 1 described above, or may include filament images acquired from cameras in other spinning and drawing devices 1.

[0125] <Functional Configuration of the Information Processing Device 100>

[0126] Next, with reference to Figures 8 to 12 , the functional configuration of the information processing device 100 will be described. Figure 8 is a diagram showing an example of the functional configuration of the information processing device 100.

[0127] The information processing device 100 executes the above-described learning program 126 or the above-described anomaly detection program 128, and thereby functions as a preprocessing unit 151, a learning unit 152, a preprocessing unit 153, a determination unit 154, and an output unit 155. As an example, the functions of the preprocessing unit 151 and the learning unit 152 are installed in the learning program 126, and the functions of the preprocessing unit 153, the determination unit 154, and the output unit 155 are installed in the above-described anomaly detection program 128.

[0128] Hereinafter, the preprocessing unit 151, the learning unit 152, the preprocessing unit 153, the determination unit 154, and the output unit 155 will be described in order.

[0129] (F1. Preprocessing Unit 151)

[0130] First, with reference to Figure 9 and Figure 10 , the function of the preprocessing unit 151 shown in Figure 8 will be described. Figure 9 is a diagram for explaining an example of the preprocessing performed by the preprocessing unit 151. Figure 10 is a diagram for explaining another example of the preprocessing performed by the preprocessing unit 151.

[0131] The preprocessing unit 151 performs preprocessing on the filament images specified in the above-described learning data 123 (see Figure 7 ). As a result, the filament images are processed into a form suitable for detecting filament path anomalies.

[0132] In Figure 9 In the example shown, the preprocessing unit 151 performs preprocessing for extracting edges from the filament image IM1 for learning. More specifically, the preprocessing unit 151 applies an edge filter (i.e., a differential filter) to the filament image IM1 for learning, and generates a filament image IM2 in which the edge portions are emphasized.

[0133] The background portion in the filament image IM1 is fuzzier than the oil supply guide 11 and the filament F. Therefore, the background portion disappears through the edge extraction process, and the contours of the oil supply guide 11 and the filament F remain in the filament image IM2. As a result, the information unnecessary for the detection of the filament path abnormality disappears. Consequently, the detection accuracy of the filament path abnormality is improved.

[0134] Preferably, the preprocessing unit 151 attaches image information AD1 and AD2 to the filament image IM2 after edge extraction. In Figure 9 the example, image information AD1 of a stripe pattern is attached to the upper part of the filament image IM2. On the other hand, image information AD2 of a stripe pattern is attached to the lower part of the filament image IM2. By attaching the image information AD1 and AD2 to the filament image IM2, it is possible to prevent the learning of the learning unit 152 described later from ending prematurely.

[0135] In Figure 10 the example shown, the preprocessing unit 151 performs preprocessing for removing the portion of the filament image IM1 that reflects the oil supply guide 11. The method for removing the portion that reflects the oil supply guide 11 is not particularly limited.

[0136] In a certain case, the preprocessing unit 151 searches for the image region R2 that reflects the oil supply guide 11 by searching for the oil supply guide 11 from the filament image IM1, and removes the image region R2 from the filament image IM1.

[0137] Various existing image processes are used in the search process of the oil supply guide 11. As an example, a learned model is used to identify the oil supply guide 11 in the filament image IM1. The learned model is pre-generated through a learning process using a learning dataset. The learning dataset includes a plurality of learning images that reflect the oil supply guide 11. Each learning image is associated with a label indicating whether the oil supply guide 11 is reflected. The internal parameters of the learned model are pre-optimized through a learning process using such a learning dataset.

[0138] The learning method for generating a learned model can adopt various machine learning algorithms. As an example, the machine learning algorithm can adopt deep learning, convolutional neural network (CNN), fully convolutional neural network (FCN), support vector machine, etc.

[0139] In another case, the preprocessing unit 151 determines a predetermined range within the filament image IM1 as the image region R2 in which the oil supply guide 11 is reflected, and removes the image region R2 from the filament image IM1. Typically, the predetermined range is the lower region within the filament image IM1.

[0140] After that, the preprocessing unit 151 removes the image region R2 from the filament image IM1 for learning, and generates the remaining image region R1 as the filament image IM3 for learning. As Figure 10 shown, only the filaments F passing through positions upstream of the oil supply guide 11 are reflected in the filament image IM3 for learning, and the oil supply guide 11 is not reflected.

[0141] By using the filament image IM3 from which the oil supply guide 11 has been removed for the learning process described later, the information processing device 100 can detect filament path abnormalities without being affected by the type of the oil supply guide 11. As a result, the detection accuracy of filament path abnormalities is improved. In addition, the designer does not need to collect filament images for learning for all types of oil supply guides 11, and the detection function of filament path abnormalities can be more easily realized.

[0142] Preferably, the preprocessing unit 151 adds image information AD3 and AD4 to the filament image IM3 after removing the image region R2. In Figure 10 the example, image information AD3 with a stripe pattern is added to the upper part of the filament image IM3. On the other hand, image information AD4 with a stripe pattern is added to the lower part of the filament image IM3. By adding the image information AD3 and AD4 to the filament image IM3, it is possible to prevent the learning of the learning unit 152 described later from ending prematurely.

[0143] It should be noted that the preprocessing performed by the preprocessing unit 151 may not necessarily be executed. For example, when the above camera 30 is configured such that the oil supply guide 11 is not included in the shooting range, only multiple filaments F passing through positions upstream of the oil supply guide 11 are reflected in the filament image IM for learning. In this case, it is not necessary to execute Figure 10 the preprocessing shown.

[0144] In addition, the preprocessing unit 151 can execute Figure 9 the preprocessing shown andFigure 10 Both of the preprocessings shown can also perform the preprocessing of only one party.

[0145] In addition, the preprocessing unit 151 can also perform Figure 9 and Figure 10 a preprocessing different from the preprocessing shown. As an example, the preprocessing unit 151 can also perform a process of extracting a portion of the filament image IM1 that reflects the oil supply guide 11 and the filament F. In other words, the preprocessing unit 151 can also perform a process of removing portions other than the portion of the filament image IM1 that reflects the oil supply guide 11 and the filament F. For example, the preprocessing unit 151 removes portions other than the oil supply guide 11 and the filament F by extracting a predetermined area within the filament image IM1.

[0146] (F2. Learning Unit 152)

[0147] Next, the functions of the learning unit 152 shown in Figure 11 are described with reference to Figure 8 for illustration. Figure 11 is a diagram for explaining an example of the learning process performed by the learning unit 152.

[0148] The learning unit 152 performs a learning process for generating the autoencoder 124A. The autoencoder 124A is an example of the above-mentioned estimation model 124. The autoencoder 124A machine-learns to restore the filament image after compressing a normal filament image in which no filament path abnormality has occurred.

[0149] The machine learning algorithm used in the learning process is not particularly limited. For example, a neural network such as deep learning can be used. Hereinafter, the learning process using a neural network is described.

[0150] As shown in Figure 11 , the autoencoder 124A is composed of an input layer X, an intermediate layer H, and an output layer Y.

[0151] The input layer X is configured to receive the input of the normal filament image preprocessed by the preprocessing unit 151. The input layer X is composed of, for example, N units x1 to x N (where N is a natural number). The number of units constituting the input layer X is the same as the dimension of the input normal filament image.

[0152] As an example, when the number of pixels in the normal filament image is N pixels and each pixel of the normal filament image is directly input to the input layer X, the input layer X is composed of N units. As another example, the feature amount extracted from the normal filament image features can also be input to the input layer X. In this case, the input layer X is configured such that the number of its units is the same as the dimension of the feature amount. Each unit constituting the input layer X outputs the input data to each unit of the first layer of the intermediate layer H.

[0153] The intermediate layer H is composed of one or more layers. In Figure 11 the example, the intermediate layer H is composed of L layers (L is a natural number). Each layer of the intermediate layer H contains multiple units. The number of units in each layer of the intermediate layer H can be the same or different. In Figure 11 the example, the first layer of the intermediate layer H is composed of Q units h A1 ~h AQ (Q is a natural number). In addition, the final layer of the intermediate layer H is composed of R units h L1 ~h LR (R is a natural number).

[0154] Each unit constituting each layer of the intermediate layer H is connected to each unit of the previous layer and each unit of the next layer. Each unit of each layer receives the output value from each unit of the previous layer, multiplies each output value by a weight, accumulates the results of these multiplication operations, adds (or subtracts) a specified deviation to the accumulated result, inputs the result of this addition operation (or subtraction operation) to a specified function (for example, the Sigmoid function), and outputs the output value of this function to each unit of the next layer.

[0155] In the autoencoder 124A, the number of units in each layer constituting the intermediate layer H is less than the number of units constituting the input layer X. As a result, the dimension of the normal filament image is compressed during the transfer from the input layer X to the intermediate layer H.

[0156] The output layer Y is used to restore the normal filament image compressed by the intermediate layer H. More specifically, the output layer Y is composed of the same number of units as the input layer X. As an example, when the input layer X is composed of N units, the output layer Y is composed of N units. In Figure 11 the example, the output layer Y is composed of N units y1~y N . Hereinafter, the units y1~y N will also be referred to as unit y.

[0157] Each unit y is connected to each unit h L1 ~h LRConnection. Each unit y receives the output values of the units in the final layer of the intermediate layer H, multiplies each output value by a weight, accumulates the results of these multiplication operations, adds (or subtracts) a specified deviation to the accumulated result, inputs the result of this addition operation (or subtraction operation) to a specified function (e.g., Sigmoid function), and outputs the output result of this function as the output value.

[0158] Next, the update process of the internal parameters of the autoencoder 124A performed by the learning unit 152 will be described.

[0159] The learning unit 152 inputs the pixels P(1) to P(N) of the first normal filament image to the autoencoder 124A. Thus, the autoencoder 124A compresses the first normal filament image. Then, the autoencoder 124A restores the compressed filament image in a manner close to the input normal filament image. The dimension of the restored filament image is the same as that of the input normal filament image. That is, the restored filament image is composed of pixels P′(1) to P′(N). Next, the learning unit 152 calculates the error "Z" between the input filament image and the restored filament image. As an example, the error "Z" is calculated based on the following formula (1).

[0160] Z = {(P(1) - P′(1)) 2 + ··· + (P(N) - P′(N)) 2} / N ……(1)

[0161] Next, the learning unit 152 updates the internal parameters (e.g., weights, biases) of the autoencoder 124A to make the error "Z" smaller. The update of the internal parameters is achieved, for example, by the error backpropagation method.

[0162] The learning unit 152 repeatedly performs the update process of the internal parameters of the autoencoder 124A for each normal filament image to be learned. Thus, the autoencoder 124A learns to restore the normal filament image after compressing it.

[0163] That is, the autoencoder 124A outputs a filament image similar to the input normal filament image when a normal filament image is input, and outputs a filament image different from the input abnormal filament image when an abnormal filament image is input. In other words, the autoencoder 124A functions as a kind of filter that allows normal filament images to pass through but does not allow abnormal filament images to pass through.

[0164] (F3. Preprocessing unit 153)

[0165] Next, Figure 8 the function of the preprocessing unit 153 shown will be described.

[0166] The preprocessing unit 153 performs preprocessing on the filament image for estimation obtained from the above-described camera 30 (see Figure 1 ). Typically, the preprocessing unit 153 performs the same preprocessing as the above-described preprocessing unit 151.

[0167] As an example, the preprocessing unit 153 performs a process of extracting edges from the filament image for estimation. The edge extraction process is the same as that described in Figure 9 , and thus its description will not be repeated.

[0168] As another example, the preprocessing unit 153 performs a process of removing the portion in the filament image for estimation that reflects the fuel supply guide 11. This removal process is the same as that described in Figure 10 , and thus its description will not be repeated.

[0169] It should be noted that the preprocessing performed by the preprocessing unit 153 may not necessarily be executed. For example, when the above-described camera 30 is configured such that the fuel supply guide 11 is not included in the shooting range, only multiple filaments F passing through a position upstream of the fuel supply guide 11 are reflected in the filament image for estimation. In this case, the above-described removal process is not required.

[0170] In addition, the preprocessing unit 153 may perform both the above-described edge extraction process and the above-described removal process, or may perform only one of the preprocessings.

[0171] In addition, the preprocessing unit 153 may also perform preprocessing different from the above-described edge extraction process and the above-described removal process. As an example, the preprocessing unit 153 may also perform a process of cutting out the portion in the filament image IM1 that reflects the fuel supply guide 11 and the filaments F. In other words, the preprocessing unit 153 may also perform a process of removing the portion other than the portion in the filament image IM1 that reflects the fuel supply guide 11 and the filaments F. For example, the preprocessing unit 153 removes the portion other than the fuel supply guide 11 and the filaments F by cutting out a predetermined area within the filament image IM1.

[0172] (F4. Judgment unit 154)

[0173] Next, with reference to Figure 12 , the function of the judgment unit 154 shown in Figure 8 will be described. Figure 12 is a diagram for explaining an example of the judgment process performed by the judgment unit 154.

[0174] The judgment unit 154 judges whether a filament path abnormality of the filament F has occurred based on the similarity between the preprocessed filament image for estimation by the above-described preprocessing unit 153 and the restored filament image obtained by inputting the filament image for estimation to the autoencoder 124A.

[0175] More specifically, first, the determination unit 154 acquires the autoencoder 124A and inputs the filament image for estimation, which has been preprocessed by the preprocessing unit 153, into the autoencoder 124A. The acquisition destination of the autoencoder 124A can be a storage device within the information processing device 100 or an external device. When a normal filament image is input, the autoencoder 124A outputs a filament image similar to the normal filament image, and when an abnormal filament image is input, the autoencoder 124A outputs a filament image different from the abnormal filament image.

[0176] After that, the determination unit 154 calculates the similarity between the filament image for estimation and the restored filament image. In the calculation of the similarity, any algorithm can be adopted for the calculation method of this similarity. As an example, the calculation algorithm of this similarity can adopt Mean Squared Error (MSE), Sum of Squared Difference (SSD), Sum of Absolute Difference (SAD), Normalized Cross-Correlation (NCC), or Zero-mean Normalized Cross-Correlation (ZNCC), etc.

[0177] The determination unit 154 determines whether a filament path abnormality has occurred based on the comparison result between the calculated similarity and a preset threshold. Here, the magnitude relationship of the calculated similarity can vary depending on the adopted algorithm. That is, regarding the value of the calculated similarity, sometimes the more similar the filament image for estimation and the restored filament image are to each other, the smaller the similarity is, and sometimes the more similar the filament image for estimation and the restored filament image are to each other, the larger the similarity is.

[0178] As an example, an algorithm is adopted in which the more similar the filament image for estimation and the restored filament image are to each other, the smaller the similarity is. In this case, the determination unit 154 determines that a filament path abnormality has occurred when the calculated similarity is equal to or greater than the preset threshold. On the other hand, the determination unit 154 determines that no filament path abnormality has occurred when the calculated similarity is less than the preset threshold.

[0179] As another example, an algorithm is adopted in which the more similar the filament image for estimation and the restored filament image are to each other, the larger the similarity is. In this case, the determination unit 154 determines that a filament path abnormality has occurred when the calculated similarity is less than or equal to the preset threshold. On the other hand, the determination unit 154 determines that no filament path abnormality has occurred when the calculated similarity is greater than the preset threshold.

[0180] The above-mentioned threshold value can also be arbitrarily set by the setter or user of the information processing device 100. Preferably, the above-mentioned threshold value is set in advance. Since the detection accuracy of the thread path abnormality depends on the setting of the threshold value, the detection accuracy of the thread path abnormality is improved by determining the threshold value in advance.

[0181] As described above, in the present embodiment, the determination unit 154 can detect the thread path abnormality by using the autoencoder 124A that has been learned using the normal filament image. That is, in the present embodiment, the designer can implement the detection function of the thread path abnormality without collecting various learning data indicating the thread path abnormality.

[0182] (F5. Output unit 155)

[0183] Next, Figure 8 the function of the output unit 155 shown will be described.

[0184] The output unit 155 outputs a control instruction corresponding to the determination result of the determination unit 154 to a specified output destination.

[0185] In a certain case, the control instruction from the output unit 155 is output to the above-mentioned display 106 (refer to Figure 6 ). Thereby, the display 106 displays a warning indicating that a thread path abnormality has occurred. Preferably, an abnormal filament image is also displayed on the display 106.

[0186] In another case, the control instruction from the output unit 155 is output to a notification lamp (not shown) provided in the spinning and drawing device 1. As an example, when the determination result of the determination unit 154 indicates normal, the output unit 155 lights the notification lamp in a specific color (for example, green). On the other hand, when the determination result of the determination unit 154 indicates a thread path abnormality, the output unit 155 lights the notification lamp in a color different from that in the normal state (for example, red).

[0187] In yet another case, the control instruction from the output unit 155 is output to a notification buzzer (not shown) provided in the spinning and drawing device 1. As an example, when the determination result of the determination unit 154 indicates normal, the output unit 155 does not sound the notification buzzer. On the other hand, when the determination result of the determination unit 154 indicates a thread path abnormality, the output unit 155 sounds the notification buzzer in a predetermined manner.

[0188] It should be noted that in the above, an example in which the output result of the output unit 155 is output to the equipment inside the spinning and drawing device 1 has been described. However, the output result of the output unit 155 can also be sent to an external device different from the spinning and drawing device 1. As an example, this output result can also be sent to a pre-registered communication terminal. Thus, the person in charge and the manager can identify that a filament path abnormality has occurred inside the spinning and drawing device 1.

[0189] <G. Configuration mode>

[0190] As described above, the camera 30 is arranged inside the spinning and drawing device 1 so as to be able to photograph a plurality of filaments F passing through the oil supply guide 11. And, the light source 40 is arranged inside the spinning and drawing device 1 so as to be able to irradiate a plurality of filaments F passing through the oil supply guide 11.

[0191] At this time, by irradiating the filaments F with the light source 40, the brightness difference between the filaments F and the background becomes clear, and it is easy to remove noise such as the background in the edge detection processing in the above-mentioned preprocessing units 151 and 153. As a result, the detection accuracy of the filament path abnormality of the filaments F is improved.

[0192] Preferably, the camera 30 and the light source 40 are arranged so that the optical axis of the light source 40 is inclined with respect to the optical axis of the camera 30. Thereby, the amount of light reflected toward the camera 30 can be reduced. As a result, the generation of whitening in the filament image can be prevented.

[0193] More preferably, the light source 40 is a condenser illumination that can locally irradiate a specific place. The irradiation angle of the light of this condenser illumination is, for example, within 45°. Preferably, the irradiation angle of the light of this condenser illumination is, for example, within 30°. This condenser illumination is arranged so that at least a plurality of filaments passing through the oil supply guide 11 are included within its illumination range 40R. That is, this condenser illumination is arranged so as not to irradiate objects other than the filaments as much as possible. In addition, this condenser illumination is arranged so that its illumination range 40R intersects the optical axis of the camera 30.

[0194] Thereby, the irradiation of light to objects other than the filaments can be suppressed. Therefore, the amount of light reflected toward the camera 30 can be further reduced. As a result, the generation of whitening in the filament image can be more reliably prevented.

[0195] Hereinafter, with reference to Figures 13 to 15 , a specific example will be described regarding the arrangement mode of the camera 30 and the light source 40 with respect to the oil supply guide 11.

[0196] (G1. Specific example 1)

[0197] Figure 13 is a diagram for explaining an example of the arrangement mode of the camera 30 and the light source 40. In Figure 13In [the figure], the positional relationship among the fuel supply guide 11, the camera 30, and the light source 40 is shown from the right side.

[0198] In this example, the light source 40 is arranged to overlap with the camera 30 when viewed from above or below. In addition, the light source 40 is arranged at a position lower than the camera 30. Thereby, the amount of light reflected toward the camera 30 can be reduced. As a result, the generation of whitening in the filament image can be prevented.

[0199] In Figure 13 's example, the optical axis 40AX of the light source 40 intersects with the optical axis 30AX of the camera 30. The angle θ formed by the optical axes 30AX and 40AX is greater than 0° and less than 90°. The angle θ can be 30° or more, can be 45° or more, and can also be 60° or more.

[0200] In addition, the light source 40 is arranged on the front side (i.e., the front direction) closer to the viewer than the fuel supply guide 11 in the direction of observing the fuel supply guide 11 from the camera 30 (i.e., the rear direction). As a result, the amount of light directly incident from the light source 40 into the camera 30 is reduced. Thereby, the generation of whitening in the filament image can be prevented.

[0201] Preferably, an antireflection member 60 for preventing the reflection of light irradiated from the light source 40 is provided inside the spinning and drawing device 1. The antireflection member 60 is arranged on the farther side (i.e., the rear direction) from the fuel supply guide 11 in the direction of observing the fuel supply guide 11 from the camera 30. In this case, when viewed from the front side, the respective mechanisms are arranged in the order of "camera 30 (light source 40) → fuel supply guide 11 → antireflection member 60". Thereby, the antireflection member 60 prevents the light irradiated from the light source 40 from being reflected to the camera 30. As a result, the generation of whitening in the filament image can be more reliably prevented.

[0202] More preferably, the antireflection member 60 is black. Thereby, the reflectivity of black is lower than that of other colors. Therefore, the black antireflection member 60 can more reliably prevent the light irradiated from the light source 40 from being reflected to the camera 30.

[0203] (G2. Specific Example 2)

[0204] Figure 14 is a diagram for explaining another example of the arrangement mode of the camera 30 and the light source 40. In Figure 14 In [the figure], the positional relationship among the fuel supply guide 11, the camera 30, and the light source 40 is shown from the right side.

[0205] The light source 40 is arranged to overlap with the camera 30 when viewed from above or below. In this example, the light source 40 is arranged at a position above the camera 30. Thereby, the amount of light reflected toward the camera 30 can be reduced. As a result, the generation of whitening in the filament image can be prevented.

[0206] In Figure 14 the example, the optical axis 40AX of the light source 40 intersects with the optical axis 30AX of the camera 30. The angle θ formed by the optical axes 30AX and 40AX is greater than 0° and less than 90°. The angle θ can be 30° or more, can be 45° or more, and can also be 60° or more.

[0207] In addition, the light source 40 is arranged on the front side (i.e., the front direction) closer to the camera 30 than the oil supply guide 11 in the direction of observing the oil supply guide 11 from the camera 30 (i.e., the rear direction). As a result, the amount of light directly incident on the camera 30 from the light source 40 is reduced. Thereby, the generation of whitening in the filament image can be prevented.

[0208] Preferably, an antireflection member 60 for preventing the reflection of light irradiated from the light source 40 is provided inside the spinning and drawing device 1. Since the antireflection member 60 is as described above, its description will not be repeated.

[0209] In addition, in the above, an example in which one light source 40 is arranged has been described, but the number of light sources 40 can also be two or more. As an example, the first light source 40 can be arranged above the camera 30, and the second light source 40 can be arranged below the camera 30.

[0210] (G3. Specific Example 3)

[0211] Figure 15 is a diagram for explaining another example of the arrangement mode of the camera 30 and the light source 40. In Figure 15 it, the positional relationship among the oil supply guide 11, the camera 30, and the light source 40 is shown from above.

[0212] In this example, the light source 40 is arranged on the right side of the camera 30. Thereby, the amount of light reflected toward the camera 30 can be reduced. As a result, the generation of whitening in the filament image can be prevented.

[0213] In Figure 15 the example, the optical axis 40AX of the light source 40 intersects with the optical axis 30AX of the camera 30. The angle θ formed by the optical axes 30AX and 40AX is greater than 0° and less than 90°. The angle θ can be 30° or more, can be 45° or more, and can also be 60° or more.

[0214] In addition, the light source 40 is disposed closer to the front side (i.e., the front direction) than the oil supply guide 11 in the direction of observing the oil supply guide 11 from the camera 30 (i.e., the rear direction). As a result, the amount of light directly incident from the light source 40 into the camera 30 is reduced. Thereby, the generation of whitening in the filament image can be prevented.

[0215] Preferably, an antireflection member 60 for preventing reflection of light irradiated from the light source 40 is provided inside the spinning and drawing device 1. Since the antireflection member 60 is as described above, its description will not be repeated.

[0216] In addition, in the above, an example in which the light source 40 is disposed on the right side of the camera 30 has been described, but the light source 40 may also be disposed on the left side of the camera 30.

[0217] In addition, in the above, an example in which one light source 40 is disposed has been described, but the number of light sources 40 may also be two or more. As an example, the first light source 40 may be disposed on the right side of the camera 30, and the second light source 40 may be disposed on the left side of the camera 30.

[0218] <H. Flowchart of Learning Process>

[0219] Next, a flowchart of the learning process of the information processing device 100 will be described with reference to Figure 16 FIG. Figure 16 is a flowchart showing the flow of the learning process.

[0220] The control device 101 of the information processing device 100 executes various processes shown in Figure 6 by executing the above learning program 126 (refer to Figure 16 ). In another case, Figure 16 part or all of the processes shown in

[0221] may also be executed by circuit elements or other hardware. Figure 8 In step S110, the control device 101 functions as the above-described preprocessing unit 151 (refer to

[0222] ) and performs a prescribed preprocessing on the filament image as the learning data 123. Since this preprocessing is as described above, its description will not be repeated.

[0223] In step S112, the control device 101 inputs the filament image preprocessed in step S110 to the above-described autoencoder 124A. Figure 8)Function is exerted to calculate the error between the filament image input to the autoencoder 124A and the restored filament image output from the autoencoder 124A. Thereafter, the control device 101 updates the internal parameters of the autoencoder 124A in such a manner that this error is smaller than the current one. This parameter is updated, for example, by the error backpropagation method. Since the learning process in step S114 is as described above, its description will not be repeated.

[0224] In step S120, the control device 101 determines whether to end the learning process. As an example, when the estimation accuracy using the test data exceeds the desired accuracy, the control device 101 determines that the learning process ends. Alternatively, when the number of updates of the internal parameters of the autoencoder 124A exceeds a specified number, the control device 101 determines that the learning process ends.

[0225] When the control device 101 determines to end the learning process (Yes in step S120), it ends Figure 16 the processing shown. Otherwise (No in step S120), the control device 101 returns the control to step S112.

[0226] <I. Flowchart of Abnormality Detection Processing>

[0227] Next, with reference to Figure 17 , the flowchart of the abnormality detection processing executed by the information processing device 100 will be described. Figure 17 is a flowchart showing the flow of the abnormality detection processing.

[0228] The control device 101 of the information processing device 100 executes Figure 6 ) to execute Figure 17 the various processes shown. In another case, Figure 17 part or all of the processes shown may also be executed by circuit elements or other hardware.

[0229] In step S210, the control device 101 acquires a filament image for estimation from the above-described camera 30.

[0230] In step S212, the control device 101 functions as the above-described preprocessing unit 153 (refer to Figure 8 ) and executes prescribed preprocessing on the filament image acquired in step S210. Since this preprocessing is as described above, its description will not be repeated.

[0231] In step S214, the control device 101 inputs the preprocessed filament image in step S212 to the autoencoder 124A that has completed learning.

[0232] In step S216, the control device 101 functions as the determination unit 154 described above (refer to Figure 8 ) and calculates the similarity between the filament image input to the autoencoder 124A and the restored filament image output from the autoencoder 124A.

[0233] In step S230, the control device 101 functions as the determination unit 154 described above and determines whether the similarity calculated in step S216 satisfies an abnormal condition. This abnormal condition is satisfied when the filament image and the restored filament image are not similar. As an example, in the case of an algorithm where the more similar the estimated filament image and the restored filament image are to each other, the smaller the similarity is, the above abnormal condition is satisfied when the similarity exceeds a specified threshold.

[0234] When the control device 101 determines that the similarity calculated in step S216 satisfies the abnormal condition (Yes in step S230), it switches the control to step S232. Otherwise (No in step S230), the control device 101 switches the control to step S234.

[0235] In step S232, the control device 101 functions as the output unit 155 described above (refer to Figure 8 ) and outputs a determination result indicating that a filament path abnormality has occurred. Since the output process is as described above, its description will not be repeated.

[0236] In step S234, the control device 101 functions as the output unit 155 described above and outputs a determination result indicating normality. Since the output process is as described above, its description will not be repeated.

[0237] Preferably, Figure 17 the abnormal detection process shown is periodically executed when the spinning and drawing device 1 manufactures a filament.

[0238] <J. First Modification Example>

[0239] Next, with reference to Figure 18 , a first modification example of the above-described embodiment will be described. Figure 18 is a diagram showing an example of the device configuration of the filament abnormality detection system 500 in this modification example.

[0240] In the above Figure 8 example, the learning function and the abnormal detection function are installed in the same information processing device 100. However, the learning function and the abnormal detection function do not necessarily need to be installed in the same information processing device 100. As an example, the learning function and the abnormal detection function can also be installed in different information processing devices 100.

[0241] As Figure 18As shown, the filament abnormality detection system 500 includes one or more spinning and drawing devices 1 and one or more information processing devices 100. In Figure 18 In the example of, the filament abnormality detection system 500 is composed of three spinning and drawing devices 1A to 1C and two information processing devices 100A and 100B.

[0242] The information processing device 100A collects the above-mentioned learning data 123 (refer to Figure 7 ) from the spinning and drawing devices 1 (for example, spinning and drawing devices 1A and 1B) connected to the network NW. Then, the learning unit 152 of the information processing device 100A performs a learning process using the learning data 123 that has been preprocessed by the preprocessing unit 151, and generates the above-mentioned estimation model 124. The generated estimation model 124 is sent to the information processing device 100B.

[0243] The information processing device 100B obtains the filament image for estimation from the spinning and drawing device 1C. Then, the determination unit 154 of the information processing device 100B inputs the filament image for estimation that has been preprocessed by the preprocessing unit 153 into the autoencoder 124A. Then, the determination unit 154 of the information processing device 100B determines whether a filament path abnormality has occurred based on the similarity between the filament image for estimation and the restored filament image obtained from the autoencoder 124A. Since this determination process is as described above, its description will not be repeated. After that, the output unit 155 of the information processing device 100B outputs the determination result to the spinning and drawing device 1C.

[0244] <K. Second Variant Example>

[0245] In the above, the estimation model 124 as the autoencoder 124A has been described, but the estimation model 124 is not limited to the autoencoder 124A.

[0246] Hereinafter, with reference to Figures 19 to 21 , an example of detecting a filament path abnormality using an estimation model 124 other than the autoencoder 124A will be described.

[0247] (K1. Learning Dataset 122)

[0248] First, with reference to Figure 19 , the learning dataset 122 for generating an estimation model 124 other than the autoencoder 124A will be described. Figure 19 is a diagram showing the learning dataset 122 of this variant example.

[0249] The learning dataset 122 contains a plurality of learning data 123. The number of learning data 123 contained in the learning dataset 122 is arbitrary. As an example, the number of learning data 123 is several tens to several hundreds of thousands.

[0250] The learning data 123 includes learning filament images that map multiple filaments passing through the oil supply guide 11. In this modification example, a label indicating whether a filament path abnormality has occurred is further associated with each learning filament image. In other respects, it is the same as the Figure 7 learning data 123 shown.

[0251] As an example, the labels defined in the learning data set 122 include "filament path abnormality" indicating that the filament has deviated from the normal filament path on the oil supply guide 11 and "normal" indicating that the filament has passed through the normal filament path on the oil supply guide 11. Each label can be distinguished by a combination of numerical values or by a combination of character strings.

[0252] (K2. Learning unit 152)

[0253] Next, refer to Figure 20 to Figure 8 a modification example of the learning unit 152 shown. Figure 20 is a diagram for explaining an example of the learning process performed by the learning unit 152.

[0254] The learning unit 152 generates the estimation model 124B of this modification example through a learning process using the Figure 19 learning data set 122 shown. The machine learning algorithm used in the learning process is not particularly limited. For example, deep learning, convolutional neural network (CNN), fully convolutional neural network (FCN), R-CNN (Regions with Convolutional Neural Network), Fast R-CNN, Faster R-CNN, YOLO (You Only Look Once), support vector machine, etc. can be used. Hereinafter, the learning process using deep learning will be described.

[0255] As Figure 20 shown, the estimation model 124B is composed of an input layer X, an intermediate layer H, and an output layer Y.

[0256] The input layer X is configured to receive a filament image as the learning data 123 as an input. The input layer X is composed of, for example, N units x1 to x N (N is a natural number). The number of units constituting the input layer X is the same as the dimension of the input information.

[0257] As an example, when the number of pixels of the filament image used for learning is N pixels, and each pixel of the filament image is directly input to the input layer X, the input layer X is composed of N units. As another example, feature extraction processing (preprocessing) can also be performed on the filament image. In this case, the feature quantity extracted by the feature is input to the input layer X. The input layer X is constituted so that the number of its units is the same as the dimension of the feature quantity after feature extraction. Each unit constituting the input layer X outputs the input data to each unit of the first layer of the intermediate layer H.

[0258] The middle layer H is composed of one or more layers. Figure 20 In the example, the intermediate layer H is composed of L layers (L is a natural number). Each layer of the intermediate layer H contains a plurality of units. The number of units in each layer of the intermediate layer H can be the same or different. Figure 20 In the example, the first layer of the intermediate layer H consists of Q units h A1 ~h AQ (Q is a natural number). In addition, the final layer of the intermediate layer H consists of R units h L1 ~h LR Composition (R is a natural number).

[0259] Each unit of each layer constituting the intermediate layer H is connected to each unit of the previous layer and each unit of the next layer. Each unit of each layer receives an output value from each unit of the previous layer, multiplies each output value by a weight, accumulates the multiplication results, adds (or subtracts) a prescribed deviation from the accumulated result, inputs the addition result (or subtraction result) to a prescribed function (e.g., Sigmoid function), and outputs the output value of the function to each unit of the next layer.

[0260] The output layer Y outputs an estimation result corresponding to the inputted filament image. The output layer Y is composed of, for example, units y1 and y2.

[0261] Units y1 and y2 are respectively connected to the final unit h of the intermediate layer H. L1 ~h LR Connection. Units y1 and y2 receive the output values of each unit of the final layer from the intermediate layer H, multiply each output value by a weight, accumulate the multiplication results, add (or subtract) a specified deviation from the accumulated result, input the addition result (or subtraction result) into a specified function (for example, a Sigmoid function), and output the output result of the function as an output value.

[0262] The number of units constituting the output layer Y is determined according to the number of categories of the labels specified in the learning data 123. As an example, when the labels specified in the learning data 123 are "abnormal filament path" and "normal", the number of units constituting the output layer Y is two units, y1 and y2. In this case, unit y1 outputs a score "sa" indicating the possibility of an abnormal filament path occurring. Unit y2 outputs a score "sb" indicating the possibility of the filament path of the filament being normal.

[0263] Next, an explanation will be given of the update process of the internal parameters of the estimation model 124B performed by the learning unit 152.

[0264] First, the learning unit 152 inputs the filament image specified in the first learning data 123 into the estimation model 124B. Next, the learning unit 152 compares the estimation results "sa", "sb" output from the estimation model 124B with the correct scores "sa′", "sb′" corresponding to the label associated with the first learning data 123.

[0265] As an example, when the label associated with the learning data 123 is "abnormal filament path", the correct score is (sa′, sb′) = (1, 0). On the other hand, when the label associated with the learning data 123 is "normal", the correct score is (sa′, sb′) = (0, 1).

[0266] The learning unit 152 calculates the error "Z" between the output results "sa", "sb" of the estimation model 124B and the correct scores "sa′", "sb′". The error "Z" is calculated, for example, based on the following formula (2).

[0267] Z = {(sa - sa′) 2 + (sb - sb′) 2} / 2...(2)

[0268] Next, the learning unit 152 updates various parameters (such as weights and biases) included in the estimation model 124B to make the error "Z" smaller. The update of this parameter is achieved, for example, by the error backpropagation method.

[0269] The learning unit 152 repeatedly performs the update process of the internal parameters of the estimation model 124B for each learning data 123 included in the learning data set 122. As a result, the estimation model 124B outputs accurate estimation results as the learning progresses.

[0270] Note that the learning unit 152 does not need to use all the learning data 123 included in the learning data set 122 for learning processing, and a part of the learning data 123 included in the learning data set 122 can also be used to generate the estimation model 124B. The remaining learning data 123 is used, for example, for the evaluation of the estimation model 124B, etc.

[0271] (K3. Judgment unit 154)

[0272] Next, with reference to Figure 21 , for Figure 8 the modified example of the judgment unit 154 shown is described. Figure 21 is a diagram for explaining an example of the judgment process performed by the judgment unit 154.

[0273] First, the judgment unit 154 acquires the estimation model 124B generated by the learning unit 152 from the storage destination. The acquisition destination of the estimation model 124B can be the above-mentioned auxiliary storage device 120 or an external device.

[0274] Next, the judgment unit 154 causes the above-mentioned camera 30 to photograph the oil supply guide 11 through which the filament passes, and acquires a filament image for estimation from the camera 30. Next, the judgment unit 154 judges whether a filament path abnormality has occurred based on the output result obtained from the estimation model 124B by inputting the filament image for estimation into the estimation model 124B.

[0275] The estimation model 124B includes, for example, a score "sa" indicating the possibility of a filament path abnormality and a score "sb" indicating the possibility of a normal filament path. In this case, the judgment unit 154 judges that a filament path abnormality has occurred when the score "sa" exceeds the first threshold and the score "sb" is lower than the second threshold. Otherwise, the judgment unit 154 judges that no filament path abnormality has occurred.

[0276] The first threshold and the second threshold can be set in advance or can be arbitrarily set by the user. In addition, the first threshold and the second threshold can be the same or different.

[0277] Note that in the above, it has been described on the premise that the estimation model 124B outputs two scores "sa" and "sb", but the estimation model 124B can also be configured to output only the score "sa" indicating the possibility of the occurrence of a filament path abnormality. In this case, the judgment unit 154 judges that a filament path abnormality has occurred when the score "sa" exceeds the first threshold. On the other hand, the judgment unit 154 judges that no filament path abnormality has occurred when the score "sa" is below the first threshold.

[0278] The embodiments disclosed herein should be considered illustrative in all respects and not restrictive. The scope of the present invention is shown not by the above description but by the claims, and is intended to include all modifications within the meaning and scope equivalent to the claims.

Claims

1. A filament abnormality detection device that can detect abnormalities in a spinning and drawing device that draws multiple filaments spun from a spinning device, wherein, the spinning and drawing device includes a guiding member that guides the multiple filaments conveyed from the upstream side to the downstream side in a manner that brings them closer to each other, the filament abnormality detection device includes a control device, the control device performs the following processes: An acquisition process of obtaining a filament image for estimation that reflects the multiple filaments from a camera configured to include at least the multiple filaments passing through the guiding member within the shooting range; A determination process of determining whether the filament path abnormality has occurred based on a determination model that has undergone machine learning to detect the filament path abnormality in which at least one of the multiple filaments detaches from the guiding member and the filament image for estimation; And An output process of outputting the determination result in the determination process.

2. The filament abnormality detection device according to claim 1, wherein, the determination model is an autoencoder that has undergone machine learning to restore the filament image after compressing a normal filament image in which the filament path abnormality has not occurred, in the determination process, based on the similarity between the filament image for estimation and the restored filament image obtained by inputting the filament image for estimation into the autoencoder, it is determined whether the filament path abnormality has occurred.

3. The filament abnormality detection device according to claim 2, wherein, in the determination process, based on the comparison result between the similarity and a predetermined threshold value, it is determined whether the filament path abnormality has occurred.

4. The filament abnormality detection device according to claim 1, wherein, the determination model is generated by performing machine learning on a plurality of learning data, in each of the plurality of learning data, a label indicating whether the filament path abnormality has occurred is associated with a learning filament image that reflects the multiple filaments passing through the guiding member, in the determination process, based on the output result obtained from the determination model by inputting the filament image for estimation into the determination model, it is determined whether the filament path abnormality has occurred.

5. The filament abnormality detection device according to any one of claims 1 to 4, wherein, the determination process includes: A process of performing preprocessing on the filament image for estimation; and A process of inputting the preprocessed filament image for estimation into the determination model, the preprocessing includes a process of extracting edges from the filament image for estimation.

6. The filament abnormality detection device according to any one of claims 1 to 5, wherein, the spinning and drawing device further includes a light source configured to include at least the multiple filaments passing through a position upstream of the guiding member within the illumination range.

7. The filament abnormality detection device according to claim 6, wherein, the guiding member includes: A main body having a surface in contact with the multiple filaments conveyed from the spinning device in the direction of gravity; A discharge port formed on the surface for discharging the sizing agent; and Two filament guiding members are provided on the surface in such a manner as to be located on both sides in the horizontal direction of the discharge port, and the two filament guiding members are configured to guide the plurality of filaments conveyed in the direction of gravity toward the central side in the horizontal direction of the discharge port. The camera is configured to include, within the imaging range of the camera, at least the plurality of filaments passing through a position upstream of the surface. The light source is configured to include, within the illumination range of the light source, at least the plurality of filaments passing through a position upstream of the surface.

8. The filament abnormality detection device according to claim 6 or 7, wherein The optical axis of the light source is inclined with respect to the optical axis of the camera.

9. The filament abnormality detection device according to any one of claims 6 to 8, wherein The light source is provided at a position closer to the front side than the guide member in the direction of observing the guide member from the camera.

10. The filament abnormality detection device according to any one of claims 6 to 9, wherein The spinning and drawing device further includes an antireflection member for preventing reflection of light irradiated from the light source, The antireflection member is provided at a position farther from the guide member than the guide member in the direction of observing the guide member from the camera.

11. The filament abnormality detection device according to claim 10, wherein The antireflection member is black.

12. The filament abnormality detection device according to any one of claims 6 to 11, wherein The light source is a condenser illumination, The illumination range of the condenser illumination intersects with the optical axis of the camera.

13. An abnormality detection program capable of detecting an abnormality in a spinning and drawing device that draws a plurality of filaments spun from a spinning device, wherein The spinning and drawing device includes a guide member for guiding the plurality of filaments closer to each other, The abnormality detection program causes a filament abnormality detection device to perform the following steps: An acquisition step of acquiring a filament image for estimation that reflects the plurality of filaments from a camera configured to include, within the imaging range, at least the plurality of filaments passing through the guide member; An input step of inputting the filament image for estimation into a estimation model that has been machine-learned to estimate a filament path abnormality in which at least one of the plurality of filaments detaches from the guide member; A determination step of determining whether the filament path abnormality has occurred based on the output result of the estimation model; And An output step of outputting the determination result in the determination step.

14. The abnormality detection program according to claim 13, wherein The estimation model is an autoencoder that has been machine-learned to restore a normal filament image that has not had the filament path abnormality after compression, In the determination step, it is determined whether the filament path abnormality has occurred based on the similarity between the filament image for estimation and the restored filament image obtained by inputting the filament image for estimation into the autoencoder.

15. The abnormality detection program according to claim 13 or 14, wherein The determination step includes: A step of performing preprocessing on the filament image for estimation; and The step of inputting the preprocessed filament image for estimation into the estimation model, The preprocessing includes a process of extracting edges from the filament image for estimation.

Images