Tension abnormality detection device, tension abnormality detection method, and tension abnormality detection program

Abstract

To provide a tension abnormality detection device capable of accurately, in a simple manner, detecting various abnormalities related to the tension of yarn wound by a yarn winder.SOLUTION: The tension abnormality detection device includes: a controller; and a tension sensor for detecting tension of yarn wound by a yarn winder. The controller performs: a process of acquiring a self-encoder that has learned to restore, after compressing a normal tension waveform, the normal tension waveform; and a process of detecting abnormalities in the tension of yarn wound onto the yarn winder based on a degree of similarity between an input tension waveform obtained from the tension sensor and an output tension waveform obtained by inputting the input tension waveform into the self-encoder.SELECTED DRAWING: Figure 6

Description

[Technical field]

[0001] The present disclosure relates to a tension abnormality detection device, a tension abnormality detection method, and a tension abnormality detection program. [Background technology]

[0002] Patent Document 1 discloses a method for detecting an abnormality in a process in which a yarn spun from a spinning device is wound up by a winding device. A tension measuring means is provided in the yarn path. The tension measuring means measures the tension of the yarn and obtains an average tension value and a standard deviation of the tension over a predetermined sampling period. The method of Patent Document 1 compares the obtained average tension value with an upper threshold value and a lower threshold value corresponding thereto, and further compares the obtained standard deviation of the tension with an upper threshold value corresponding thereto to detect the presence or absence of an abnormality. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2002-302824 A Summary of the Invention [Problem to be solved by the invention]

[0004] In the method of Patent Document 1, abnormalities are uniformly detected based on the average tension value and the standard deviation of the tension. However, since there are various patterns of tension abnormalities, it is not easy to accurately detect various tension abnormalities based only on the average tension value and the standard deviation of the tension.

[0005] The present disclosure has been made in consideration of the above circumstances, and its main objective is to provide a tension abnormality detection device, a tension abnormality detection method, and a tension abnormality detection program that are capable of accurately detecting various abnormalities related to the tension of the yarn wound by the yarn winding machine. [Means for solving the problem]

[0006] In one example of the present disclosure, a tension abnormality detection device is provided. The tension abnormality detection device includes a control device and a tension sensor for detecting the tension of a yarn wound by a yarn winding machine. The control device executes a process of acquiring an autoencoder that has been trained to compress a normal tension waveform and then restore the normal tension waveform, and a process of detecting an abnormality in the tension of the yarn wound by the yarn winding machine based on a similarity between an input tension waveform obtained from the tension sensor and an output tension waveform obtained by inputting the input tension waveform to the autoencoder.

[0007] In this tension abnormality detection device, an autoencoder that has been trained based on a normal tension waveform is used to detect abnormalities in the tension of the thread. Therefore, various abnormalities that deviate from the normal tension can be detected. For example, abnormalities that cannot be detected only by the magnitude or degree of change of the tension can be detected.

[0008] In one example of the present disclosure, the control device further executes a learning process to generate the autoencoder using a normal tension waveform.

[0009] As a result, the learning process is executed in the tension abnormality detection device. By providing the tension abnormality detection device with a learning function, the tension abnormality detection device can generate an autoencoder by itself.

[0010] In one example of the present disclosure, the yarn winding machine is configured to wind the yarn onto a bobbin while traversing the yarn, and the tension sensor is configured to intermittently contact the yarn as the yarn is traversed to detect the tension.

[0011] As a result, the data detected by the tension sensor that intermittently contacts the yarn contains not only the maximum tension value but also a change trend that indicates how the tension changes over time. This results in the tension waveform containing many features, so an autoencoder is generated that can accurately detect the degree of abnormality in the yarn tension.

[0012] In one example of the present disclosure, the detection process detects an abnormality in tension when the similarity is equal to or greater than a predetermined threshold, and detects a normality in tension when the similarity is less than the predetermined threshold.

[0013] This allows the tension abnormality detection device to detect a tension abnormality according to the threshold value.

[0014] In one example of the present disclosure, the tension abnormality detection device further includes a display unit. The control device further executes a process of creating a graph showing a time change of the similarity and displaying it on the display unit.

[0015] This allows, for example, a manager or operator to intuitively grasp the degree of abnormality in the tension.

[0016] In one example of the present disclosure, the control device executes the detection process while the yarn winding machine is winding the yarn.

[0017] This allows the tension abnormality detection device to detect the degree of abnormality in tension in real time or close to real time, allowing the manager or operator to take action such as interrupting winding when the tension is abnormal.

[0018] In one example of the present disclosure, the control device further executes a process of receiving an input of a tension waveform and acquiring an estimation model trained to output the type of the abnormality, and, if a tension abnormality is detected in the detection process, a process of determining the type of the abnormality based on the input tension waveform and the estimation model.

[0019] As a result, the tension abnormality detection device can not only detect the occurrence of a tension abnormality, but also determine the type of tension abnormality that has occurred.

[0020] In another example of the present disclosure, a tension abnormality detection method is provided that is executed by a tension abnormality detection device. The tension abnormality detection device includes a tension sensor for detecting the tension of a yarn wound by a yarn winding machine. The tension abnormality detection method includes a step of acquiring an autoencoder that is trained to compress a normal tension waveform and then restore the normal tension waveform, and a step of detecting an abnormality in the tension of the yarn wound by the yarn winding machine based on a similarity between an input tension waveform obtained from the tension sensor and an output tension waveform obtained by inputting the input tension waveform to the autoencoder.

[0021] In this tension abnormality detection method, an autoencoder trained based on a normal tension waveform is used to detect abnormalities in the tension of the thread. Therefore, various abnormalities that deviate from the normal tension can be detected. For example, abnormalities that cannot be detected only by the magnitude or degree of change of the tension can be detected.

[0022] In another example of the present disclosure, a tension abnormality detection program is provided that is executed by a tension abnormality detection device. The tension abnormality detection device includes a tension sensor for detecting the tension of a yarn wound by a yarn winding machine. The tension abnormality detection program causes the tension abnormality detection device to execute a step of acquiring an autoencoder that has been trained to compress a normal tension waveform and then restore the normal tension waveform, and a step of detecting an abnormality in the tension of the yarn wound by the yarn winding machine based on a similarity between an input tension waveform obtained from the tension sensor and an output tension waveform obtained by inputting the input tension waveform to the autoencoder.

[0023] In this tension abnormality detection program, an autoencoder trained based on normal tension waveforms is used to detect abnormalities in the tension of the thread. Therefore, various abnormalities that deviate from the normal tension can be detected. For example, abnormalities that cannot be detected only by the magnitude or degree of change of the tension can be detected. [Brief description of the drawings]

[0024] [Figure 1]FIG. 2 is a front view of the yarn winding machine according to the embodiment. [Diagram 2] FIG. 2 is a side view of the yarn winding machine. [Diagram 3] FIG. 2 is a block diagram of the yarn winding machine. [Figure 4] FIG. 13 is a diagram showing a method for obtaining an output tension waveform from an input tension waveform using a conversion model. [Diagram 5] FIG. 13 is a graph showing an input tension waveform, an output tension waveform, and a score when there is no tension abnormality. [Figure 6] FIG. 13 is a graph showing an input tension waveform, an output tension waveform, and a score when a tension abnormality occurs. [Figure 7] FIG. 13 is a flowchart showing a process for determining whether or not there is a tension abnormality. [Figure 8] FIG. 11 is a flowchart showing a process for allowing an operator to determine whether or not there is a tension abnormality. [Figure 9] FIG. 2 is a diagram illustrating an example of a hardware configuration of a control device. [Figure 10] FIG. 1 is a diagram illustrating an example of a learning dataset. [Figure 11] FIG. 2 is a diagram illustrating an example of a functional configuration of a control device. [Figure 12] FIG. 2 is a diagram conceptually illustrating a learning process performed by a learning unit. [Figure 13] 10 is a diagram conceptually illustrating an abnormality detection process performed by a detection unit. FIG. [Figure 14] FIG. 1 illustrates an example of a device configuration of an anomaly detection system. [Figure 15] FIG. 1 is a diagram illustrating an example of a learning dataset. [Figure 16] FIG. 2 is a diagram illustrating an example of a functional configuration of a control device. [Figure 17] FIG. 2 is a diagram conceptually illustrating a learning process performed by a learning unit. [Figure 18] 10 is a diagram conceptually illustrating a process of determining an abnormality type by a determination unit. FIG. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0025] Hereinafter, each embodiment according to the present invention will be described with reference to the drawings. In the following description, the same parts and components are denoted by the same reference numerals. Their names and functions are also the same. Therefore, detailed descriptions thereof will not be repeated. Note that each embodiment and each modification described below may be selectively combined as appropriate.

[0026] <A. Bobbin winder 1> Next, an embodiment of the present disclosure will be described with reference to the drawings. FIG. 1 is a front view of a bobbin winder 1 according to an embodiment of the present disclosure. FIG. 2 is a side view of the bobbin winder 1. FIG. 3 is a block diagram of the bobbin winder 1. In the following description, the upstream or downstream in the running direction of the yarn may be simply referred to as upstream or downstream.

[0027] An unwinder (not shown) is arranged upstream of the bobbin winder 1 shown in FIG. 1. The unwinder manufactures yarn 93 and supplies it to the bobbin winder 1. The bobbin winder 1 winds the yarn 93 around a bobbin 91 to manufacture a package 94. As shown in FIG. 1, the yarn 93 is wound around the bobbin 91 at the upper winding position to form a yarn layer and a package 94 is formed. On the other hand, the yarn 93 is not wound around the bobbin 91 at the lower standby position.

[0028] The yarn 93 is a synthetic fiber yarn having excellent stretchability such as nylon or polyester. More specifically, the yarn 93 is, for example, FDY (Fully Draw Yarn) or POY (Partially Oriented Yarn). However, the type of the yarn 93 is not limited to these, and the tension abnormality detection device 10 described below can also be applied to yarns other than the types described above.

[0029] 2, in this embodiment, a plurality of yarns 93 aligned in the axial direction of the first bobbin holder 41 (second bobbin holder 42) are supplied from the spinning machine to the yarn winding machine 1 via a yarn feed roller 100. A plurality of bobbins 91 are provided aligned in the axial direction of the first bobbin holder 41 (second bobbin holder 42). The yarn winding machine 1 winds the plurality of yarns 93 onto the bobbins 91, respectively, to produce a plurality of packages 94.

[0030] The following describes in detail the yarn winding machine 1. As shown in Fig. 1, the yarn winding machine 1 includes a frame 11, a fulcrum guide 12, a first housing 20, a second housing 30, and a turret plate (bobbin holder moving mechanism) 40.

[0031] The frame 11 is a member that holds each component included in the yarn winding machine 1. A first housing 20 and a second housing 30 are attached to the frame 11. The first housing 20 and the second housing 30 are movable up and down relative to the frame 11.

[0032] The fulcrum guide 12 is disposed on the frame 11. The fulcrum guide 12 guides the yarn 93 supplied from the spinning machine to the yarn winding machine 1. In detail, the fulcrum guide 12 has a wall portion that comes into contact with the yarn 93, and regulates the position of the yarn 93 so that the yarn 93 does not deviate from the yarn path.

[0033] A traverse device 21 is attached to the first housing 20. The traverse device 21 traverses the yarn 93 sent downstream by reciprocating along the axial direction of a first bobbin holder 41 (described later) to the winding width of a package 94 with a traverse guide 23 (described later) engaged with the yarn 93. This traverse motion of the yarn 93 forms a yarn layer on the bobbin 91 or the package 94. As shown in FIG. 3, the traverse device 21 includes a traverse cam 22 and a traverse guide 23.

[0034] The traverse cam 22 is a roller-shaped member disposed in parallel to the bobbin 91 or the package 94. A spiral cam groove is formed on the outer circumferential surface of the traverse cam 22. The traverse cam 22 is rotated by a traverse motor 51.

[0035] The traverse motor 51 is controlled by a control device 50, which will be described later. The traverse guide 23 is a part that engages with the yarn 93. The tip of the traverse guide 23 has, for example, a substantially U-shaped guide portion, and engages with the yarn 93 by pinching the yarn 93 in the winding width direction. The base end of the traverse guide 23 is positioned in the cam groove of the traverse cam 22. With this configuration, by driving and rotating the traverse cam 22, the traverse guide 23 can be reciprocated in the winding width direction with the fulcrum guide 12 as a fulcrum.

[0036] A contact roller 31 is rotatably attached to the second housing 30. The contact roller 31 rotates while contacting the yarn layer of the package 94 with a predetermined pressure during winding of the yarn 93, thereby feeding the yarn 93 from the traverse guide 23 to the yarn layer of the package 94 and adjusting the shape of the yarn layer of the package 94. The contact roller 31 may be rotationally driven by a driving unit such as a motor.

[0037] The second housing 30 is provided with an operation panel 32. The operation panel 32 is a device that is operated by an operator. The operator issues instructions to the yarn winding machine 1 by operating the operation panel 32. Examples of instructions issued by the operator include starting winding, stopping winding, changing winding conditions, etc.

[0038] A tension sensor 13 is provided on the yarn path from the spinning machine to the traverse device 21. More specifically, the tension sensor 13 is provided downstream of the fulcrum guide 12 and upstream of the traverse device 21 in the running direction of the yarn 93. Thus, the tension sensor 13 detects the tension of the yarn 93 reciprocated by the traverse device 21. The tension sensor 13 has a pair of arms (not shown). One arm contacts the yarn 93 at one end of the traverse. The other arm contacts the yarn 93 at the other end of the traverse. That is, the tension sensor 13 intermittently contacts the traversed yarn 93. In addition, a strain gauge is provided on the arm, and the strain of the arm detected by the strain gauge is correlated with the tension of the yarn 93. Thus, the tension of the yarn 93 can be obtained based on the signal output by the strain gauge. A signal indicating the value of the tension detected by the tension sensor 13 is output to a control device 50 (described later). Furthermore, the signal output by the tension sensor 13 includes a tension waveform that indicates the change in tension over time.

[0039] The tension sensor 13 described above is merely an example, and a tension sensor 13 having a different configuration from that of this embodiment may be used. For example, the tension sensor 13 is not limited to a configuration in which it intermittently contacts the thread 93, and may be a configuration in which it continuously contacts the thread 93. Furthermore, the tension sensor 13 may detect tension using a spring or a piezoelectric element instead of a strain gauge.

[0040] 3, the yarn winding machine 1 includes a display unit 14. The display unit 14 is a display capable of displaying information. The display unit 14 is, for example, a liquid crystal display or an organic electroluminescence display. The display unit 14 displays, on a screen, an image output by a control device 50, which will be described later.

[0041] The turret plate 40 is a disk-shaped member. The turret plate 40 is rotatably attached to the frame 11. The turret plate 40 can rotate around a normal line passing through the center of the disk as a rotation axis. The turret plate 40 is rotationally driven by a turret motor 53 shown in FIG. 3. The turret motor 53 is controlled by a control device 50 described later.

[0042] The turret plate 40 is provided with a first bobbin holder 41 and a second bobbin holder 42 at two locations facing each other across the center of the disk. A plurality of bobbins 91 can be mounted on the first bobbin holder 41, lined up in the axial direction of the first bobbin holder 41. A plurality of bobbins 91 can be mounted on the second bobbin holder 42, lined up in the axial direction of the second bobbin holder 42. The positions of the first bobbin holder 41 and the second bobbin holder 42 can be changed by rotating the turret plate 40. Note that, as long as the positions of the first bobbin holder 41 and the second bobbin holder 42 can be changed, another device may be used instead of the turret plate 40.

[0043] The first bobbin holder 41 is rotatable relative to the turret plate 40 around the axial position of the first bobbin holder 41 as the center of rotation. The first bobbin holder 41 is rotationally driven by a first bobbin holder motor 54 shown in FIG. 3. Similarly, the second bobbin holder 42 is rotatable relative to the turret plate 40 around the axial position of the second bobbin holder 42 as the center of rotation. The second bobbin holder 42 is rotationally driven by a second bobbin holder motor 55 shown in FIG. 3. The first bobbin holder motor 54 and the second bobbin holder motor 55 are controlled by a control device 50 described later.

[0044] Hereinafter, the first bobbin holder 41 and the second bobbin holder 42 will be collectively referred to as the bobbin holders 41, 42. Fig. 1 shows the bobbin holders 41, 42 lined up one above the other. At this time, the position of the higher bobbin holder 41, 42 is the winding position, and the position of the lower bobbin holder 41, 42 is the standby position. The yarn winding machine 1 winds the yarn 93 onto the bobbin 91 of the bobbin holder 41, 42 that is in the winding position to produce a package 94.

[0045] Furthermore, when a predetermined amount of yarn 93 has been wound and the package 94 of the first bobbin holder 41 is full, the turret plate 40 rotates, switching the positions of the first bobbin holder 41 and the second bobbin holder 42. After that, the package 94 of the first bobbin holder 41, which is now full and in the standby position, is collected, and the yarn 93 is wound around the bobbin 91 of the second bobbin holder 42, which is in the winding position. A new bobbin 91 is attached to the first bobbin holder 41 from which the package 94 has been collected.

[0046] The control device 50 includes a control unit 50a, a learning unit 50b, a detection unit 50c, and a storage unit 50d. Specifically, the control device 50 is configured as a known computer, and includes a CPU (Central Processing Unit), a RAM (Random Access Memory), an SSD (Solid State Drive), and the like. The CPU is a type of processor. The SSD stores programs and data for controlling the yarn winding machine 1 in advance. The SSD also stores tensions input from the tension sensor 13 in chronological order. The CPU reads the programs into the RAM and executes them, thereby allowing the control device 50 to operate as the control unit 50a, the learning unit 50b, and the detection unit 50c. The SSD corresponds to the storage unit 50d. Note that instead of the SSD, a hard disk drive (HDD) or a flash memory may be used. Alternatively, a storage provided outside the control device 50 and capable of communicating with the control device 50 may be used as the storage unit.

[0047] The control unit 50a is responsible for the overall control performed by the control device 50. The control unit 50a processes data input from the outside to the control device 50 or data stored in the memory unit 50d. The control unit 50a stores the data obtained by this processing in the memory unit 50d or outputs it to the outside of the control device 50. The learning unit 50b performs a process of constructing a conversion model using machine learning. The conversion model constructed by the learning unit 50b will be described in detail later. The detection unit 50c performs a detection process of detecting anomalies based on data input from the outside to the control device 50 and the conversion model constructed by the learning unit 50b and stored in the memory unit 50d. The memory unit 50d stores data according to the processing of the control unit 50a.

[0048] 3 shows the tension abnormality detection device 10. The tension abnormality detection device 10 detects the degree of abnormality in the tension of the yarn 93 wound by the yarn winding machine 1. The tension abnormality detection device 10 includes a control device 50, a tension sensor 13, and a display unit 14. The control device 50, as a part constituting the tension abnormality detection device 10, performs the following processes.

[0049] That is, the control unit 50a outputs data to the learning unit 50b and the detection unit 50c to perform processing. The control unit 50a stores data obtained by the processing performed by the learning unit 50b and the detection unit 50c in the memory unit 50d. The control unit 50a also stores the tension waveform detected by the tension sensor 13 in the memory unit 50d. Note that further examples of objects controlled by the control unit 50a include the traverse motor 51, the turret motor 53, the first bobbin holder motor 54, and the second bobbin holder motor 55. The learning unit 50b performs machine learning based on the tension waveform stored in the memory unit 50d or the tension waveform based on the detected value of tension input from the tension sensor 13 to construct a conversion model. The detection unit 50c detects the degree of abnormality in the tension of the yarn 93 wound by the yarn winding machine 1 and determines whether or not there is an abnormality, based on data input from the outside to the control device 50 (specifically, the tension detected by the tension sensor 13) and the conversion model constructed by the learning unit 50b and stored in the memory unit 50d. The determination result obtained by the processing of the detection unit 50c is output to the outside by the control unit 50a or stored in the memory unit 50d. As described above, the memory unit 50d stores the data obtained by the processing of the control unit 50a, the learning unit 50b, and the detection unit 50c.

[0050] The conversion model is an autoencoder, and is constructed by the learning unit 50b performing the following machine learning. First, the operator or manager instructs the learning unit 50b to start machine learning. In response to this instruction, the learning unit 50b requests the tension detection value from the tension sensor 13. The tension sensor 13 outputs the tension detection value to the learning unit 50b. The learning unit 50b buffers the tension detection value input from the tension sensor 13. The process of the learning unit 50b acquiring and buffering the tension detection value is repeated. Thereafter, the learning unit 50b performs machine learning on the buffered tension detection value to construct the conversion model. Note that the tension detection value output by the tension sensor 13 based on the request of the learning unit 50b may be stored in the memory unit 50d, and the learning unit 50b may perform machine learning based on the contents stored in the memory unit 50d to construct the conversion model. The conversion model constructed by the learning unit 50b is stored in the memory unit 50d and set in the detection unit 50c. This completes the machine learning.

[0051] In this embodiment, tension sensor 13 is configured to intermittently come into contact with yarn 93, and therefore the tension waveform is triangular. Furthermore, the learning data for creating the conversion model is a tension waveform with normal tension. Only tension waveforms with normal tension are used to create the conversion model, and tension waveforms with abnormal tension are not used. Generally, abnormal tension occurs rarely, and therefore it is easy to prepare a tension waveform with normal tension.

[0052] The conversion model is constructed by performing unsupervised learning in the input layer, intermediate layer, and output layer of a multi-layer neural network using the above learning data. The learning performed here is deep learning, and since the feature values ​​are identified by the learning, input of the feature values ​​is not necessary. By setting a tension waveform with normal tension to the conversion model constructed by performing this learning, important feature values ​​contained in the tension waveform are extracted and compressed in the intermediate layer (hidden layer), and then a tension waveform restored based on the feature values ​​is generated. In other words, if the input tension waveform is a normal tension waveform, the conversion model is used to generate an output tension waveform similar to the input tension waveform. Note that the construction method of the conversion model in this embodiment is just one example, and other methods may be used.

[0053] As shown in Fig. 4, the conversion model is used by the detection unit 50c. Specifically, the detection unit 50c generates an output tension waveform based on the input tension waveform and the conversion model. Here, if the input tension waveform is a normal tension waveform, the detection unit 50c generates a tension waveform similar to the input tension waveform as the output tension waveform. If the input tension waveform is an abnormal tension waveform, the detection unit 50c generates a tension waveform different from the input tension waveform as the output tension waveform.

[0054] 5 shows a tension waveform in which no tension abnormality occurs (input tension waveform), and an output tension waveform generated by detection unit 50c based on the input tension waveform and the conversion model. Because the input tension waveform does not contain any tension abnormality, the input tension waveform and the output tension waveform are substantially the same and have a high degree of similarity.

[0055] 6 shows a tension waveform in which a tension abnormality occurs (input tension waveform), and an output tension waveform generated by detection unit 50c based on the input tension waveform and a conversion model. Because the input tension waveform contains a tension abnormality, the similarity between the input tension waveform and the output tension waveform is low in the time period in which the tension abnormality occurs (the time period indicated by the dashed oval).

[0056] The detection unit 50c compares the input tension waveform with the output tension waveform, and detects the degree of abnormality in the tension of the yarn 93 based on the similarity between the input tension waveform and the output tension waveform. Various methods for determining the degree of correlation between the two data can be used to compare the input tension waveform with the output tension waveform. In this embodiment, the Mahalanobis distance is used. The Mahalanobis distance is a virtual distance calculated taking into account the correlation, and the distance becomes smaller when the correlation is high. Since the Mahalanobis distance is well known, a detailed description is omitted. The detection unit 50c calculates the Mahalanobis distance for each time period based on the input tension waveform and the output tension waveform. Here, the correlation between the input tension waveform and the output tension waveform indicates the similarity of the waveforms. That is, if the tension is normal, the similarity between the input tension waveform and the output tension waveform will be high, and the Mahalanobis distance will be small. If the tension is abnormal, the similarity between the input tension waveform and the output tension waveform will be low, and the Mahalanobis distance will be large. Therefore, a large Mahalanobis distance indicates an abnormal degree of tension. Hereinafter, the Mahalanobis distance will be referred to as a score. As described above, a value other than the Mahalanobis distance can also be used as the score.

[0057] The input tension waveform in Figure 5 does not contain any tension abnormalities, so the score is low. On the other hand, the input tension waveform in Figure 6 contains tension abnormalities, so the score is high during the time period when tension abnormalities are included. By using the score, the similarity between the input tension waveform and the output tension waveform, in other words, the degree of tension abnormality, can be expressed as a specific numerical value.

[0058] 5 and 6, a threshold value may be set for the score. In this case, the threshold value is set in advance in the detection unit 50c. The threshold value may be obtained experimentally or empirically. The detection unit 50c may determine whether the tension is abnormal or not based on whether the score is equal to or greater than the threshold value or less than the threshold value.

[0059] In the example shown in FIG. 6, the tension is much greater during the time period when the tension abnormality occurs than during other time periods. This is one example of a tension abnormality, and there are various types of tension abnormalities. Depending on the type of tension abnormality, the manner in which the tension changes may differ from other time periods. This type of tension abnormality cannot be determined by using only the magnitude of the tension. However, by using the method of this embodiment, even this type of tension abnormality can be detected because the score is high.

[0060] Next, the flow of processing performed by the tension abnormality detection device 10 will be described mainly with reference to FIG.

[0061] The operator or manager sets the above-mentioned threshold value in advance in the detection unit 50c. After that, the operator or manager instructs the detection unit 50c to detect tension abnormalities. The detection unit 50c also determines whether it is time to detect a tension abnormality (S101), and when there is an instruction from the operator or manager, determines that the detection timing has arrived.

[0062] Next, the detection unit 50c generates an output tension waveform (S102). Specifically, when the detection unit 50c determines that it is detection timing, it requests a tension detection value from the tension sensor 13. The tension sensor 13 outputs the tension detection value to the detection unit 50c. The detection unit 50c buffers the tension detection value input from the tension sensor 13. The detection unit 50c then generates an output tension waveform based on an input tension waveform based on the buffered tension detection value and the conversion model stored in the memory unit 50d. Next, the detection unit 50c compares the input tension waveform with the output tension waveform and calculates a score as described above (S103).

[0063] Next, the detection unit 50c determines whether the score is equal to or greater than a threshold (S104). If the detection unit 50c determines that the score is equal to or greater than the threshold, it determines that the tension is abnormal (S105). In this case, the control unit 50a or the detection unit 50c may store the time when the tension was abnormal, etc. Furthermore, the control unit 50a or the detection unit 50c may notify the operator that the tension is abnormal by displaying on the display unit 14. Alternatively, the control unit 50a may instruct the winding of the yarn 93 to be interrupted. On the other hand, if the detection unit 50c determines that the score is less than the threshold, it determines that the tension is normal (S106).

[0064] In the example shown in Fig. 7, the detection unit 50c judges whether the tension is abnormal. Alternatively, the detection unit 50c may simply present a score as shown in Fig. 8. In this case, the operator or manager judges whether the tension is abnormal by looking at the score presented by the detection unit 50c.

[0065] 8, steps S201 to S203 are the same as steps S101 to S103 described above, and therefore will not be described. After calculating the score, the detection unit 50c displays the score numerically and graphically on the display unit 14 (S204).

[0066] 7 and 8, the degree of abnormality in the tension of the yarn 93 being wound is detected while the yarn winding machine 1 is winding the yarn 93. In other words, the degree of abnormality in the tension is detected in real time. Alternatively, data related to the tension may be stored while the yarn 93 is being wound, and the degree of abnormality in the tension may be detected based on the stored data after the yarn 93 has been wound.

[0067] As described above, the tension abnormality detection device 10 of the present embodiment detects the degree of abnormality in the tension of the yarn 93 wound by the yarn winding machine 1. The tension abnormality detection device 10 includes a tension sensor 13, a learning unit 50b, a storage unit 50d, and a detection unit 50c. The tension sensor 13 detects the tension of the yarn 93 wound by the yarn winding machine 1. The learning unit 50b uses an autoencoder to learn a tension waveform with normal tension by machine learning, thereby constructing a conversion model that generates the same waveform based on a tension waveform with normal tension. The storage unit 50d stores the conversion model constructed by the learning unit 50b. The detection unit 50c generates an output tension waveform based on an input tension waveform based on the tension detected by the tension sensor 13 and the conversion model that has learned the tension waveform, and detects the degree of abnormality in the tension of the yarn 93 wound by the yarn winding machine 1 based on the similarity between the input tension waveform and the output tension waveform.

[0068] Since the conversion model is created by learning normal tension waveforms, it can detect various abnormalities that deviate from normal tension. In particular, by constructing a conversion model through machine learning of tension waveforms using an autoencoder, it can detect abnormalities that cannot be detected by the magnitude of tension or the degree of change alone.

[0069] In the tension abnormality detection device 10 of the present embodiment, the yarn winding machine 1 traverses the yarn 93 while winding it into a package 94. The tension sensor 13 intermittently contacts the yarn 93 as the yarn 93 traverses to detect the tension.

[0070] The data detected by the tension sensor 13, which is in intermittent contact with the yarn 93, includes not only the maximum tension value but also a change trend that indicates how the tension changes over time. Therefore, the tension waveform includes many features, making it possible to create a model for detecting the degree of abnormality in the tension of the yarn 93 with high accuracy.

[0071] In the tension abnormality detection device 10 of the present embodiment, the detection unit 50c compares the input tension waveform with the output tension waveform, and calculates a score that quantifies the degree of abnormality.

[0072] This makes it possible to specifically obtain the degree of abnormality in tension.

[0073] In the tension abnormality detection device 10 of this embodiment, a threshold value is set in advance in the detection unit 50c, and the detection unit 50c determines that the tension is abnormal if the score is equal to or greater than the threshold value, and determines that the tension is normal if the score is less than the threshold value.

[0074] This makes it possible to determine whether the tension is abnormal or not.

[0075] The tension abnormality detection device 10 of this embodiment includes a display unit 14, and the detection unit 50c performs control to create and display on the display unit 14 a graph showing the change over time in the score.

[0076] This allows, for example, a manager or operator to intuitively grasp the degree of abnormality in the tension.

[0077] In the tension abnormality detection device 10 of the present embodiment, the detection section 50c detects the degree of abnormality in the tension of the yarn 93 being wound while the yarn winding machine 1 is winding the yarn 93.

[0078] This makes it possible to detect the degree of abnormality in tension in real time or at a timing close to that, and therefore, for example, when the tension is abnormal, it is possible to take action such as interrupting winding.

[0079] Although the preferred embodiment of the present disclosure has been described above, the above configuration can be modified, for example, as follows.

[0080] The flowcharts shown in the above embodiment are merely examples, and some processes may be omitted, the contents of some processes may be changed, or new processes may be added.

[0081] Although the traverse device 21 of the above embodiment is of a cam drum type, it may have a different configuration as long as it can reciprocate the traverse guide 23 in the width direction of the winding. For example, instead of the traverse device 21, a rotary traverse device using rotary blades or a belt-type traverse device that reciprocally drives the traverse guide by a belt can also be used.

[0082] In the above embodiment, an example in which the present invention is applied to a yarn winder that winds the yarn produced by a spinning machine has been described. However, instead of this yarn winder, the present invention can also be applied to a false twisting machine or a rewinding machine.

[0083] <B. Hardware Configuration of Control Device 50> Next, with reference to FIG. 9, the hardware configuration of the control device 50 shown in FIG. 3 will be described. FIG. 9 is a diagram showing an example of the hardware configuration of the control device 50.

[0084] The control device 50 includes the above-described storage unit 50d (see FIG. 3), a processor 101, a communication interface 104, a display interface 105, and an input interface 107. These components are connected to a bus 115. Examples of the storage unit 50d include a ROM (Read Only Memory) 102, a RAM 103, and an auxiliary storage device 120.

[0085] The processor 101 is constituted by, for example, at least one integrated circuit. The integrated circuit can be constituted by, for example, at least one CPU, 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.

[0086] The processor 101 controls the operation of the control device 50 by executing various programs. Based on receiving an execution command for various programs, the processor 101 reads 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 the execution of the program.

[0087] A LAN (Local Area Network), an antenna, etc. are connected to the communication interface 104. The control device 50 exchanges data with an external device via the communication interface 104. The external device includes, for example, a server.

[0088] The display interface 105 is connected to the above-mentioned display unit 14 (see FIG. 3). The display interface 105 sends an image signal for displaying an image to the display unit 14 in accordance with a command from the processor 101 or the like. The display unit 14 is, for example, a liquid crystal display, an organic EL (Electro Luminescence) display, or other displays. The display unit 14 may be configured integrally with the control device 50, or may be configured separately from the control device 50.

[0089] 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 device capable of accepting a user's operation. The input device 108 may be configured integrally with the control device 50, or may be configured separately from the control device 50.

[0090] The auxiliary storage device 120 is, for example, a storage medium such as a hard disk, a flash memory, or an SSD. The auxiliary storage device 120 stores, for example, a learning dataset 122, the above-mentioned conversion model 124, a learning program 126, and a tension abnormality detection program 128. The storage location of these is not limited to the auxiliary storage device 120, and may be stored in a memory area (for example, a cache memory, etc.) of the processor 101, the ROM 102, the RAM 103, an external device (for example, a server), etc.

[0091] The learning program 126 is a program for generating the conversion model 124 using the learning dataset 122. The learning program 126 may be provided not as a standalone program but as part of an arbitrary program. In this case, the learning process by the learning program 126 is realized in cooperation with the arbitrary program. Even if the program does not include such a part of the module, it does not deviate from the purpose of the learning program 126 according to the present embodiment. Furthermore, some or all of the functions provided by the learning program 126 may be realized by dedicated hardware. Furthermore, the control device 50 may be configured in a form like a so-called cloud service in which at least one server executes part of the processing of the learning program 126.

[0092] The tension abnormality detection program 128 is a program for detecting an abnormality in the tension of the thread wound around the thread winder 1 using the learned conversion model 124. The tension abnormality detection program 128 may be provided not as a single program but incorporated into a part of any program. In this case, the abnormality detection process by the tension abnormality detection program 128 is realized in cooperation with any program. Even a program that does not include such a part of the module does not deviate from the gist of the tension abnormality detection program 128 according to the present embodiment. Furthermore, part or all of the functions provided by the tension abnormality detection program 128 may be realized by dedicated hardware. Furthermore, the control device 50 may be configured in a form such as a so-called cloud service in which at least one server executes a part of the process of the tension abnormality detection program 128.

[0093] <C. Learning dataset 122> Next, with reference to FIG. 10, the learning dataset 122 shown in FIG. 9 will be described. FIG. 10 is a diagram showing an example of the learning dataset 122.

[0094] 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.

[0095] In each of the learning data 123, a data ID (Identification) and a tension waveform with normal tension are associated. The data ID is information for uniquely identifying the learning data 123. The data ID is input by the user so as not to overlap, for example.

[0096] The tension waveform defined in the learning data 123 is during the thread winder 1 winding the thread It is a data group in which the tensions detected by the above-described tension sensor 13 are arranged in time series. That is, in one learning data 123, the tension is associated with each time. The number of dimensions of each tension waveform defined in the learning data set 122 is equal to each other. Further, the learning data set 122 is composed only of the learning data 123 of the normal tension waveform.

[0097] <D. Functional Configuration of Control Device 50> Next, with reference to FIGS. 11 to 13, the functional configuration of the control device 50 will be described. FIG. 11 is a diagram showing an example of the functional configuration of the control device 50.

[0098] As shown in FIG. 11, the control device 50 includes a learning unit 50b and a detection unit 50c as functional configurations. Hereinafter, these functional configurations will be described in order.

[0099] (D1. Learning Unit 50b) First, with reference to FIG. 12, the function of the learning unit 50b shown in FIG. 11 will be described. FIG. 12 is a diagram conceptually showing the learning process by the learning unit 50b.

[0100] The learning unit 50b executes a learning process using the above-described learning data set 122 (see FIG. 10) and generates a conversion model 124 as an autoencoder. The machine learning algorithm adopted in the learning process is not particularly limited, and for example, a neural network such as deep learning can be adopted. Hereinafter, the learning process using a neural network will be described.

[0101] As shown in FIG. 12, the conversion model 124 is composed of an input layer X, an intermediate layer H, and an output layer Y.

[0102] The input layer X is configured to receive the input of the normal tension waveform defined in the learning data 123. The input layer X includes, for example, N units x1 to x N(N is a natural number). The number of units making up the input layer X is the same as the number of dimensions of the input tension waveform. For example, if the input tension waveform is N-dimensional data, the input layer X is made up of N units. Each unit making up the input layer X outputs the input data to each unit in the first layer of the hidden layer H.

[0103] The middle layer H is composed of one layer or multiple layers. In the example of FIG. 12, the middle layer H is composed of L layers (L is a natural number). Each layer of the middle layer H includes multiple units. The number of units in each layer of the middle layer H may be the same or different. In the example of FIG. 12, the first layer of the middle layer H is composed of Q units h A1 ~h AQ (Q is a natural number). The final layer of the hidden layer H is composed of R units h L1 ~h LR (R is a natural number).

[0104] Each unit constituting each layer of the intermediate layer H is connected to each unit in the previous layer and each unit in the next layer. Each unit in each layer receives an output value from each unit in the previous layer, multiplies each output value by a weight, accumulates the multiplication results, adds (or subtracts) a predetermined bias to (from) the accumulated result, inputs the addition result (or subtraction result) to a predetermined function (e.g., the Sigmonite function), and outputs the output value of the function to each unit in the next layer.

[0105] In the transformation model 124 as an autoencoder, the number of units constituting each layer of the hidden layer H is smaller than the number of units constituting the input layer X. As a result, the number of dimensions of the tension waveform is compressed in the process of being transmitted from the input layer X to the hidden layer H.

[0106] The output layer Y is configured to restore the tension waveform compressed in 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 the example of FIG. 12, the output layer Y is composed of N units y1 to yN Hereinafter, units y1 to y3 will also be referred to as unit y.

[0107] Each unit y is connected to each unit h L1 ~h LM Each of the units y receives an output value from each unit in the final layer of the hidden layer H, multiplies each output value by a weight, accumulates the results of these multiplications, adds (or subtracts) a predetermined bias to (from) the accumulated result, inputs the result of the addition (or subtraction) to a predetermined function (e.g., the Sigmonite function), and outputs the output result of the function as an output value.

[0108] Next, the update process of the internal parameters of the conversion model 124 by the learning unit 50b will be described.

[0109] The learning unit 50b inputs the normal tension waveform T(t) defined in the first learning data 123 to the conversion model 124. As a result, the conversion model 124 compresses the tension waveform T(t) and restores the tension waveform T(t) to a tension waveform T'(t) of the same dimension. Next, the learning unit 50b calculates the error "Z" between the input tension waveform T(t) and the output tension waveform T'(t). As an example, the error "Z" is calculated based on the following equation (1).

[0110] Z = {(T(t1)-T'(t1)) 2 +···+(T(t N )-T'(t N )) 2} / N···(1) Next, the learning unit 50b updates the internal parameters (for example, weights and biases) of the conversion model 124 so as to reduce the error "Z." The update of the internal parameters is realized, for example, by the error backpropagation method.

[0111] The learning unit 50b repeatedly performs the process of updating the internal parameters of the conversion model 124 for each piece of learning data 123 included in the learning data set 122. In this way, the conversion model 124 learns to restore a normal tension waveform after compressing the normal tension waveform. In other words, when a normal tension waveform is input, the conversion model 124 comes to output a tension waveform similar to the normal tension waveform, and when an abnormal tension waveform is input, the conversion model 124 comes to output a tension waveform different from the abnormal tension waveform. In other words, the conversion model 124 functions like a kind of filter that passes a normal tension waveform while not passing an abnormal normal tension waveform.

[0112] The learning unit 50b does not need to use all of the learning data 123 included in the learning dataset 122 for the learning process, and may generate the conversion model 124 using a portion of the learning data 123 included in the learning dataset 122. The remaining learning data 123 is used, for example, for evaluation of the conversion model 124.

[0113] (D2. Detection unit 50c) Next, the function of the detector 50c shown in Fig. 11 will be described with reference to Fig. 13. Fig. 13 is a diagram conceptually showing an abnormality detection process by the detector 50c.

[0114] The detection unit 50c detects an abnormality in the tension of the yarn being wound by the yarn winding machine 1 based on the similarity between the input tension waveform obtained from the tension sensor 13 and the output tension waveform obtained by inputting the input tension waveform to the conversion model 124.

[0115] More specifically, first, the detection unit 50c inputs the input tension waveform obtained from the tension sensor 13 into the conversion model 124, and acquires the output tension waveform from the conversion model 124. The acquisition destination of the conversion model 124 may be the above-described storage unit 50d or an external device. When a normal tension waveform is input, the conversion model 124 outputs a tension waveform similar to the normal tension waveform, and when an abnormal tension waveform is input, the conversion model 124 outputs a tension waveform different from the abnormal tension waveform. The detection unit 50c detects an abnormality in the tension of the thread wound around the winder 1 based on the degree of similarity between the input tension waveform and the output tension waveform. Since the function of the detection unit 50c is as described above, a detailed description of its function will not be repeated.

[0116] <E. Modified Example 1> Next, with reference to FIG. 14, a modified example related to the functional arrangement of the winder 1 will be described.

[0117] In the example of FIG. 11 described above, the learning unit 50b and the detection unit 50c were implemented in the same winder 1. However, the learning unit 50b and the detection unit 50c do not necessarily have to be implemented in the same winder 1. As an example, the learning unit 50b may be implemented in a different device.

[0118] FIG. 14 is a diagram showing an example of the device configuration of the abnormality detection system 500 in this modified example. As shown in FIG. 14, the abnormality detection system 500 includes one or more winders 1 and one or more information processing devices 200. In the example of FIG. 14, the abnormality detection system 500 is composed of three winders 1A to 1C and one information processing device 200.

[0119] The winders 1A to 1C and the information processing device 200 are connected to the same network NW and are configured to be able to communicate with each other. The winder 1 and the information processing device 200 may be communicatively connected by wire or wirelessly.

[0120] The information processing device 200 is a desktop personal computer, a notebook personal computer, a tablet terminal, or other information processing terminal.

[0121] In the example of FIG. 14, the learning unit 50b is implemented in the information processing device 200. Further, the detection unit 50c is implemented in the bobbin winder 1C.

[0122] The information processing device 200 collects the above-described learning data 123 (see FIG. 10) from the bobbin winders 1 (for example, bobbin winders 1A and 1B) connected to the network NW. Next, the learning unit 50b of the information processing device 200 executes a learning process using the learning data 123 collected from all the bobbin winders 1, and generates the above-described conversion model 124. Since the learning function of the learning unit 50b is as described above, the description thereof will not be repeated.

[0123] Thereafter, the information processing device 200 transmits the generated conversion model 124 to the bobbin winder 1 (for example, bobbin winder 1C). The detection unit 50c of the bobbin winder 1C uses the conversion model 124 to detect a tension abnormality occurring in the bobbin winder 1C. Since the abnormality detection function of the detection unit 50c is as described above, the description thereof will not be repeated.

[0124] <F. Modification Example 2>

[0125] (F1. Overview) Next, with reference to FIGS. 15 to 18, the tension abnormality detection device 10 according to the modification example will be described.

[0126] In the above example, the tension abnormality detection device 10 determines whether or not a tension abnormality has occurred by inputting the input tension waveform obtained from the tension sensor 13 into the conversion model 124. On the other hand, the tension abnormality detection device 10 according to this modification example further discriminates the type of the occurring tension abnormality when it is determined that a tension abnormality has occurred. Note that since other points such as the hardware configuration of the tension abnormality detection device 10 are as described above, the descriptions thereof will not be repeated.

[0127] (F2. Training Dataset 132) Next, the learning data set 132 used when generating an estimation model for discriminating the type of tension abnormality will be described with reference to Fig. 15. Fig. 15 is a diagram showing an example of the learning data set 132.

[0128] The training data set 132 includes a plurality of training data 133. The number of training data 133 included in the training data set 132 is arbitrary. As an example, the number of training data 133 is several tens to several hundreds of thousands.

[0129] A data ID (Identification), a tension waveform with abnormal tension, and a type of tension abnormality are associated with each piece of learning data 133. The data ID is information for uniquely identifying the learning data 133. The data ID is input by the user, for example, so as not to be duplicated.

[0130] The tension waveform defined in the learning data 133 is obtained while the yarn winding machine 1 is winding the yarn. This is a data group in which tensions detected by the above-mentioned tension sensor 13 are arranged in chronological order. The number of dimensions of each tension waveform defined in learning dataset 132 is equal to each other. Furthermore, each tension waveform is associated with an abnormality type as a label. The abnormality type is input by the user, for example, via the above-mentioned input device 108 or the like.

[0131] (F3. Functional configuration of the control device 50) Next, the functional configuration of the control device 50 in this modified example will be described with reference to Fig. 16 to Fig. 18. Fig. 16 is a diagram showing an example of the functional configuration of the control device 50.

[0132] The control device 50 includes, as its functional configuration, a learning unit 50b, a detection unit 50c, a learning unit 50e, and a discrimination unit 50f. The control device 50 shown in FIG. 16 differs from the control device 50 shown in FIG. 11 in that it further includes a learning unit 50e and a discrimination unit 50f. The functions of the learning unit 50b and the detection unit 50c have been described above, and therefore will not be repeated. Below, the functions of the learning unit 50e and the discrimination unit 50f will be described in order.

[0133] It is not necessary that all of the learning unit 50b, the detection unit 50c, the learning unit 50e, and the discrimination unit 50f are implemented in the control device 50. As an example, at least one of the learning unit 50b and the learning unit 50e may be implemented in the information processing device 200 described above.

[0134] (a) Learning unit 50e First, the function of the learning unit 50e shown in Fig. 16 will be described with reference to Fig. 17. Fig. 17 is a diagram conceptually showing the learning process by the learning unit 50e.

[0135] The learning unit 50e executes a learning process using the above-mentioned learning data set 132 (see FIG. 15) to generate an estimation model 134 for discriminating tension abnormalities. The machine learning algorithm employed in the learning process is not particularly limited, and various machine learning algorithms such as neural networks such as deep learning, support vector machines, or decision trees may be employed. The learning process using deep learning will be described below.

[0136] As shown in FIG. 17, the estimation model 134 is composed of an input layer X, an intermediate layer H, and an output layer Y.

[0137] The input layer X is configured to receive an abnormal tension waveform defined in the learning data 133. The input layer X includes, for example, N units x1 to x N(N is a natural number). The number of units making up the input layer X is the same as the number of dimensions of the input tension waveform. For example, if the input tension waveform is N-dimensional data, the input layer X is made up of N units. Each unit making up the input layer X outputs the input data to each unit in the first layer of the hidden layer H.

[0138] The middle layer H is composed of one layer or multiple layers. In the example of FIG. 17, the middle layer H is composed of L layers (L is a natural number). Each layer of the middle layer H includes multiple units. The number of units in each layer of the middle layer H may be the same or different. In the example of FIG. 17, the first layer of the middle layer H is composed of Q units h A1 ~h AQ (Q is a natural number). The final layer of the hidden layer H is composed of R units h L1 ~h LR (R is a natural number).

[0139] Each unit constituting each layer of the intermediate layer H is connected to each unit in the previous layer and each unit in the next layer. Each unit in each layer receives an output value from each unit in the previous layer, multiplies each output value by a weight, accumulates the multiplication results, adds (or subtracts) a predetermined bias to (from) the accumulated result, inputs the addition result (or subtraction result) to a predetermined function (e.g., the Sigmonite function), and outputs the output value of the function to each unit in the next layer.

[0140] The output layer Y outputs an estimation result according to the input tension waveform. The output layer Y is composed of units y1 to y3, for example. Hereinafter, the units y1 to y3 are also referred to as units y.

[0141] Each unit y is connected to a unit h L1 ~h LREach of the units y receives an output value from each unit in the final layer of the hidden layer H, multiplies each output value by a weight, accumulates the results of these multiplications, adds (or subtracts) a predetermined bias to (from) the accumulated result, inputs the result of the addition (or subtraction) to a predetermined function (e.g., the Sigmonite function), and outputs the output result of the function as an output value.

[0142] The number of units constituting the output layer Y is determined according to the number of types of tension abnormalities defined in the learning data 133. As an example, when detecting tension abnormalities "A" to "C", the number of units constituting the output layer Y is three, units y1 to y3. In this case, unit y1 is configured to output a score "sa" indicating the possibility that tension abnormality "A" has occurred. Unit y2 is configured to output a score "sb" indicating the possibility that tension abnormality "B" has occurred. Unit y3 is configured to output a score "sc" indicating the possibility that tension abnormality "C" has occurred.

[0143] 17, the estimation model 134 is configured to output a plurality of scores, but one estimation model may be configured to output one score. As an example, the first estimation model may be configured to output a score "sa" related to the anomaly type "A", the second estimation model may be configured to output a score "sb" related to the anomaly type "B", and the third estimation model may be configured to output a score "sc" related to the anomaly type "C".

[0144] Next, the process of updating the internal parameters of the estimation model 134 by the learning unit 50e will be described.

[0145] The learning unit 50e inputs the tension waveform T(t) defined in the first learning data 133 to the estimation model 134. Next, the learning unit 50e compares the estimation results “sa” to “sc” output from the estimation model 134 with the correct scores “sa′” to “sc′” corresponding to the abnormality types associated with the first learning data 133.

[0146] As an example, when the abnormality type associated with the learning data 133 is "A", the correct score is (sa', sb', sc') = (1, 0, 0). When the abnormality type associated with the learning data 133 is "B", the correct score is (sa', sb', sc') = (0, 1, 0). When the abnormality type associated with the learning data 133 is "C", the correct score is (sa', sb', sc') = (0, 0, 1).

[0147] The learning unit 50e calculates an error "Z" between the output results "sa" to "sc" of the estimation model 134 and the correct scores "sa'" to "sc'". The error "Z" is calculated, for example, based on the following formula (2).

[0148] Z={(sa-sa') 2 +(sb-sb') 2 +(sc-sc') 2} / 3···(2) Next, the learning unit 50e updates various parameters (for example, weights and biases) included in the estimation model 134 so as to reduce the error "Z." The updating of the parameters is realized, for example, by the backpropagation method.

[0149] The learning unit 50e repeatedly performs the update process of the internal parameters of the estimation model 134 for each piece of learning data 133 included in the learning data set 132. As a result, the estimation model 134 begins to output accurate estimation results as the learning progresses.

[0150] Note that the learning unit 50e does not need to use all of the learning data 133 included in the learning dataset 132 for the learning process, and may generate the estimation model 134 using a portion of the learning data 133 included in the learning dataset 132. The remaining learning data 133 is used, for example, for evaluation of the estimation model 134.

[0151] (b) Discrimination section 50f Next, the function of the discriminating section 50f shown in Fig. 16 will be described with reference to Fig. 18. Fig. 18 is a diagram conceptually showing the process of discriminating the type of abnormality by the discriminating section 50f.

[0152] First, the discrimination unit 50f acquires the estimation model 134 generated by the learning unit 50e. The estimation model 134 may be acquired from the storage unit 50d described above or from an external device. When a tension abnormality is detected by the detection unit 50c described above, the discrimination unit 50f discriminates the type of the occurring tension abnormality based on the input tension waveform used to detect the tension abnormality and the estimation model 134.

[0153] More specifically, discrimination unit 50f inputs an input tension waveform indicating an abnormality to estimation model 134. As a result, estimation model 134 outputs a score indicating the possibility that a tension abnormality has occurred for each abnormality type. Estimation model 134 outputs, for example, as estimation results, a score "sa" indicating the possibility that tension abnormality "A" has occurred, a score "sb" indicating the possibility that tension abnormality "B" has occurred, and a score "sa" indicating the possibility that tension abnormality "C" has occurred.

[0154] The discrimination unit 50f discriminates the type of tension abnormality that has occurred based on the scores "sa" to "sc". As an example, when any of the scores "sa" to "sc" exceeds a predetermined threshold, the discrimination unit 50f outputs the tension abnormality corresponding to the maximum score among the scores "sa" to "sc" as the discrimination result. On the other hand, when none of the scores "sa" to "sc" exceed the predetermined threshold, the discrimination unit 50f outputs that an unknown tension abnormality has occurred. The predetermined threshold may be set in advance or may be arbitrarily set by the user.

[0155] Note that the output mode of the discrimination result by the discrimination unit 50f is not particularly limited. As an example, the discrimination unit 50f outputs a warning including the type of the occurring tension abnormality. The warning may be, for example, displayed as a message on the above-described display unit 14, output as voice, or stored in the storage unit 50d as a log.

[0156] <G. Addendum> As described above, the present embodiment includes the following disclosure.

[0157] From the viewpoint of the present disclosure, a tension abnormality detection device having the following configuration is provided. That is, the tension abnormality detection device detects the degree of abnormality of the tension of the thread wound by the winder. The tension abnormality detection device includes a tension sensor, a learning unit, a storage unit, and a detection unit. The tension sensor detects the tension of the thread wound by the winder. The learning unit constructs a conversion model that generates the same waveform based on the tension waveform with normal tension by performing machine learning on the tension waveform with normal tension using an autoencoder. The storage unit stores the conversion model constructed by the learning unit. The detection unit generates an output tension waveform based on the input tension waveform based on the tension detected by the tension sensor and the conversion model, and detects the degree of abnormality of the tension of the thread wound by the winder based on the similarity between the input tension waveform and the output tension waveform.

[0158] Since the tension waveform with normal tension is learned to create a conversion model, various abnormalities deviating from normal tension can be detected. In particular, by constructing a model through machine learning of the tension waveform using an autoencoder, abnormalities that cannot be detected only by the magnitude or degree of change of the tension can be detected.

[0159] In the above-described tension abnormality detection device, it is preferable to have the following configuration. That is, the winder winds the thread onto a bobbin while traversing the thread. The tension sensor intermittently contacts the thread according to the traverse of the thread to detect the tension.

[0160] The data detected by the tension sensor that intermittently contacts the yarn includes not only the maximum tension value but also the change trend that indicates how the tension changes over time. As a result, the tension waveform contains many features, making it possible to create a model for detecting the degree of abnormality in the yarn tension with high accuracy.

[0161] In the tension abnormality detection device, it is preferable that the detection section compares the input tension waveform with the output tension waveform and calculates a score that quantifies the degree of abnormality.

[0162] This makes it possible to specifically obtain the degree of abnormality in tension.

[0163] In the above-mentioned tension abnormality detection device, it is preferable that a threshold value is set in advance in the detection unit, and the detection unit determines that the tension is abnormal if the score is equal to or greater than the threshold value, and determines that the tension is normal if the score is less than the threshold value.

[0164] This makes it possible to determine whether the tension is abnormal or not.

[0165] It is preferable that the tension abnormality detection device further comprises a display unit, and the detection unit performs control to create a graph showing the change in the score over time and display it on the display unit.

[0166] This allows, for example, a manager or operator to intuitively grasp the degree of abnormality in the tension.

[0167] In the tension abnormality detection device, it is preferable that the learning unit and the detection unit detect an abnormality in the tension of the yarn being wound while the yarn winding machine is winding the yarn.

[0168] This makes it possible to detect the degree of abnormality in tension in real time or at a timing close to that, and therefore, for example, when the tension is abnormal, it is possible to take action such as interrupting winding.

[0169] The embodiments disclosed herein should be considered to be illustrative and not restrictive in all respects. The scope of the present invention is defined by the claims, not the above description, and is intended to include all modifications within the meaning and scope of the claims. [Explanation of symbols]

[0170] 1 Yarn winding machine 10. Tension Abnormality Detection Device 14 Display section 13 Tension sensor 50 Control device 50b Learning Section 50c Detection unit 50d storage section 50f Discrimination part 124 Conversion Model 134 Estimation Model

Claims

1. Control device and It is equipped with a tension sensor for detecting the tension of the thread being wound onto the thread winding machine, The control device is A process to obtain an autoencoder that has been trained to restore a normal tension waveform after compressing it, A tension anomaly detection device that performs a process to detect an anomaly in the tension of the yarn being wound onto the yarn winding machine, based on the similarity between the input tension waveform obtained from the tension sensor and the output tension waveform obtained by inputting the input tension waveform into the autoencoder.

2. The tension anomaly detection device according to claim 1, further comprising a control device that performs a learning process to generate the autoencoder using a normal tension waveform.

3. The aforementioned thread winding machine winds the thread onto the bobbin while traversing it. The tension abnormality detection device according to claim 1, wherein the tension sensor is configured to intermittently contact the thread in accordance with the traverse of the thread to detect tension.

4. The thread winding machine winds the thread onto the bobbin while traversing, The tension abnormality detection device according to claim 2, wherein the tension sensor is configured to intermittently contact the thread in accordance with the traverse of the thread to detect tension.

5. In the aforementioned detection process, If the similarity is above a predetermined threshold, an abnormality in tension is detected. The tension abnormality detection device according to claim 1, wherein normal tension is detected when the similarity is less than the predetermined threshold.

6. In the detection process, If the similarity is above a predetermined threshold, an abnormality in tension is detected. The tension abnormality detection device according to claim 2, wherein normal tension is detected when the similarity is less than the predetermined threshold.

7. In the detection process, If the similarity is above a predetermined threshold, an abnormality in tension is detected. The tension abnormality detection device according to claim 3, wherein normal tension is detected when the similarity is less than the predetermined threshold.

8. In the detection process, If the similarity is above a predetermined threshold, an abnormality in tension is detected. The tension abnormality detection device according to claim 4, wherein normal tension is detected when the similarity is less than the predetermined threshold.

9. The tension abnormality detection device further includes a display unit, The tension anomaly detection device according to any one of claims 1 to 8, wherein the control device further performs a process of creating a graph showing the time change of the similarity and displaying it on the display unit.

10. The tension abnormality detection device according to any one of claims 1 to 8, wherein the control device performs the detection process while the thread winding machine is winding thread.

11. The tension abnormality detection device according to claim 9, wherein the control device performs the detection process while the thread winding machine is winding thread.

12. The control device further, A process to acquire an estimation model that has been trained to receive a tension waveform input and output the type of anomaly, A tension anomaly detection device according to any one of claims 1 to 8, wherein when a tension anomaly is detected in the detection process, a process is executed to determine the type of the anomaly based on the input tension waveform and the estimation model.

13. The control device further, A process to acquire an estimation model that has been trained to receive a tension waveform input and output the type of anomaly, The tension anomaly detection device according to claim 9, wherein, when a tension anomaly is detected in the detection process, a process is executed to determine the type of the anomaly based on the input tension waveform and the estimation model.

14. The control device further, A process to acquire an estimation model that has been trained to receive a tension waveform input and output the type of anomaly, The tension anomaly detection device according to claim 10, wherein when a tension anomaly is detected in the detection process, a process is executed to determine the type of the anomaly based on the input tension waveform and the estimation model.

15. The control device further, A process to acquire an estimation model that has been trained to receive a tension waveform input and output the type of anomaly, The tension anomaly detection device according to claim 11, wherein if a tension anomaly is detected in the detection process, a process is executed to determine the type of the anomaly based on the input tension waveform and the estimation model.

16. A tension anomaly detection method performed by a tension anomaly detection device, The tension abnormality detection device includes a tension sensor for detecting the tension of the thread being wound onto the thread winding machine. The tension abnormality detection method is as follows: A step of obtaining an autoencoder that has been trained to restore a normal tension waveform after compressing it, A tension anomaly detection method comprising the step of detecting an anomaly in the tension of a thread being wound onto a thread winding machine based on the similarity between an input tension waveform obtained from the tension sensor and an output tension waveform obtained by inputting the input tension waveform into the autoencoder.

17. A tension anomaly detection program executed by a tension anomaly detection device, The tension abnormality detection device includes a tension sensor for detecting the tension of the thread being wound onto the thread winding machine. The tension anomaly detection program is provided to the tension anomaly detection device. A step of obtaining an autoencoder that has been trained to restore a normal tension waveform after compressing it, A tension anomaly detection program that performs the step of detecting an anomaly in the tension of the thread being wound onto the thread winding machine based on the similarity between the input tension waveform obtained from the tension sensor and the output tension waveform obtained by inputting the input tension waveform into the autoencoder.

Images