The storage medium stores the estimation program, the information processing device, and the estimation method.

Abstract

The invention relates to a program-based estimated storage device (128) that allows estimating the probability of uneven dyeing occurring in fully drawn yarn produced by a spinning machine (3) which spins yarn while heating yarn causing the information processing device (100) to perform: the step of acquiring an autoencoder (124) trained to compress the normal tension waveform representing the normal tension transition of fully drawn yarn and then recover the normal tension waveform; the step of acquiring the input tension waveform representing the tension transition of fully drawn yarn produced by the spinning machine (3); and the step of estimating the probability of uneven dyeing occurring in fully drawn yarn produced by the spinning machine (3) based on the degree of similarity between the input tension waveform and the output tension waveform acquired from the autoencoder (124) by inputting the input tension waveform to the autoencoder (124).

Description

Estimation program, information processing device, and estimation method

[0001] The present disclosure relates to an estimation program, an information processing device, and an estimation method.

[0002] Dyeing spots may occur in fully drawn yarn (FDY) that has been subjected to a drawing process. Dyeing spots have a significant impact on the commercial value of the yarn. For this reason, dyeing tests have conventionally been carried out to determine in advance the likelihood of dyeing spots occurring. As shown in JP 2015-212924 A (Patent Document 1), in the dyeing test, a sample obtained by cylindrically knitting the yarn to be evaluated is actually dyed, and the dyeing state of the sample is evaluated.

[0003] JP 2015-212924 A

[0004] In a drawn yarn manufacturing apparatus, a dye test is performed on all packages formed by winding a drawn yarn onto a bobbin to determine the likelihood of dye spotting. That is, in the dye test, a process of actually dyeing the yarn is performed on all packages. Therefore, the dye test requires an enormous amount of man-hours. Therefore, there is a need for a technology that can estimate the likelihood of dye spotting without performing a dye test on a produced drawn yarn.

[0005] In one example of the present disclosure, there is provided an estimation program capable of estimating the possibility of dyeing irregularities occurring in a drawn yarn produced by a spinning and drawing apparatus that draws the yarn while heating it. The estimation program causes an information processing device to execute the following steps: acquiring an autoencoder that has been trained to compress a normal tension waveform that shows the normal progression of tension of a drawn yarn different from the drawn yarn and then restore the normal tension waveform; acquiring an input tension waveform that shows the progression of tension of the drawn yarn produced by the spinning and drawing apparatus; and estimating the possibility of dyeing irregularities occurring in the drawn yarn produced by the spinning and drawing apparatus based on the similarity between the input tension waveform and an output tension waveform obtained from the autoencoder by inputting the input tension waveform into the autoencoder.

[0006] This estimation program can estimate the likelihood of dye spotting occurring in a drawn yarn produced in a spinning drawing apparatus from an input tension waveform that indicates the progression of the tension of the drawn yarn. This makes it possible to estimate the likelihood of dye spotting occurring without conducting a dye test on the produced drawn yarn. As a result, the number of steps required for testing to evaluate the quality of the drawn yarn is significantly reduced.

[0007] In one example of the present disclosure, the estimation program further causes the information processing device to execute a learning step of generating the autoencoder using the normal tension waveform.

[0008] This allows the learning process and the estimation process to be performed in the same information processing device, i.e., the information processing device can generate an autoencoder within the device itself.

[0009] In one example of the present disclosure, the learning step includes a step of performing a first preprocessing on the normal tension waveform and a step of generating the autoencoder from the normal tension waveform after the first preprocessing, and the estimating step includes a step of performing a second preprocessing on the input tension waveform and a process of inputting the input tension waveform after the second preprocessing into the autoencoder.

[0010] As a result, the normal tension waveform is processed into a format suitable for learning by the first pre-processing, and the input tension waveform is processed into a format suitable for estimation by the second pre-processing.

[0011] In one example of the present disclosure, the first pre-processing includes subtracting the mean value of the normal tension waveform from the normal tension waveform, and the second pre-processing includes subtracting the mean value of the input tension waveform from the input tension waveform.

[0012] The magnitude of tension does not correlate well with the likelihood of chromophores forming. Therefore, by subtracting the average value from the normal tension waveform, features with low correlation to the likelihood of chromophores forming are removed from the normal tension waveform. Similarly, by subtracting the average value from the input tension waveform, features with low correlation to the likelihood of chromophores forming are removed from the input tension waveform. As a result, the accuracy of estimating the likelihood of chromophores forming is improved.

[0013] In one example of the present disclosure, the first pre-processing includes dividing the normal tension waveform into fixed intervals, and the second pre-processing includes dividing the input tension waveform into fixed intervals.

[0014] By dividing the normal tension waveform into fixed sections, it is possible to capture waveforms that are correlated with the likelihood of chromogenic spots occurring. In other words, feature quantities that have a low correlation with the likelihood of chromogenic spots occurring are removed from the normal tension waveform. Similarly, by dividing the input tension waveform into fixed sections, it is possible to capture waveforms that are correlated with the likelihood of chromogenic spots occurring. In other words, feature quantities that have a low correlation with the likelihood of chromogenic spots occurring are removed from the input tension waveform. As a result, the accuracy of estimating the likelihood of chromogenic spots occurring is improved.

[0015] In one example of the present disclosure, the certain section corresponds to a length of the drawn yarn of 6.0 m or less.

[0016] This allows a waveform that is more correlated with the likelihood of mottling to be captured, further improving the accuracy of estimating the likelihood of mottling.

[0017] Another example of the present disclosure provides an information processing device capable of estimating the likelihood of dye spots occurring in a drawn yarn produced by a spinning and drawing apparatus that draws the yarn while heating it. The information processing device includes a processor for controlling the information processing device. The processor executes the following processes: acquiring an autoencoder trained to compress a normal tension waveform that shows a normal progression of tension in a drawn yarn different from the drawn yarn and then restore the normal tension waveform; acquiring an input tension waveform that shows a progression of tension in the drawn yarn produced by the spinning and drawing apparatus; and estimating the likelihood of dye spots occurring in the drawn yarn produced by the spinning and drawing apparatus based on the similarity between the input tension waveform and an output tension waveform obtained from the autoencoder by inputting the input tension waveform into the autoencoder.

[0018] This information processing device can estimate the likelihood of dye spotting occurring in a drawn yarn produced in a spinning drawing apparatus from an input tension waveform that indicates the transition of tension in the drawn yarn. This makes it possible to estimate the likelihood of dye spotting occurring without conducting a dye test on the produced drawn yarn. As a result, the number of steps required for testing to evaluate the quality of the drawn yarn is significantly reduced.

[0019] In one example of the present disclosure, the input tension waveform is measured by a measurement device separate from the spinning / drawing device, which includes first and second rollers around which a yarn can be passed, a heater disposed between the first roller and the second roller for heating the yarn passed between the first roller and the second roller, and a tension sensor disposed to measure the tension of the drawn yarn traveling between the first roller and the second roller.

[0020] As a result, the input tension waveform for estimating the likelihood of dye unevenness in the drawn yarn is measured by a measuring device separate from the spinning and drawing apparatus, and therefore the likelihood of dye unevenness in the drawn yarn can be estimated without changing the device configuration of an existing spinning and drawing apparatus.

[0021] In one example of the present disclosure, the spinning and drawing apparatus further includes a pre-heating roller that heats the yarn before drawing, at least one heat setting roller that is arranged downstream of the pre-heating roller in the yarn running direction, heats the yarn at a higher temperature than the pre-heating roller, and feeds the yarn at a yarn feeding speed faster than the yarn feeding speed of the pre-heating roller, and a tension sensor that is arranged to measure the tension of the drawn yarn traveling between the pre-heating roller and the heat setting roller that is located furthest downstream in the yarn running direction among the at least one heat setting roller.

[0022] By providing a tension sensor inside the spinning drawing device, an input tension waveform for estimating the possibility of dyeing spots occurring can be acquired during the drawn yarn production process, thereby reducing the time required to estimate the possibility of dyeing spots occurring.

[0023] In one example of the present disclosure, the spinning and drawing device further includes first and second rollers over which the yarn can be passed, a heater disposed between the first roller and the second roller for heating the yarn passed over the first roller and the second roller, and a tension sensor disposed to measure the tension of the drawn yarn traveling between the first roller and the second roller.

[0024] By providing a tension sensor inside the spinning drawing device, an input tension waveform for estimating the possibility of dyeing spots occurring can be acquired during the drawn yarn production process, thereby reducing the time required to estimate the possibility of dyeing spots occurring.

[0025] In one example of the present disclosure, the processor further executes a learning process that generates the autoencoder using the normal tension waveform.

[0026] This allows the learning process and the estimation process to be performed in the same information processing device, i.e., the information processing device can generate an autoencoder within the device itself.

[0027] In one example of the present disclosure, the learning process includes a process of performing a first preprocessing on the normal tension waveform and a process of generating the autoencoder from the normal tension waveform after the first preprocessing, and the estimating process includes a process of performing a second preprocessing on the input tension waveform and a process of inputting the input tension waveform after the second preprocessing to the autoencoder.

[0028] As a result, the normal tension waveform is processed into a format suitable for learning by the first pre-processing, and the input tension waveform is processed into a format suitable for estimation by the second pre-processing.

[0029] In one example of the present disclosure, the first pre-processing includes subtracting the mean value of the normal tension waveform from the normal tension waveform, and the second pre-processing includes subtracting the mean value of the input tension waveform from the input tension waveform.

[0030] The magnitude of tension does not correlate well with the likelihood of chromophores forming. Therefore, by subtracting the average value from the normal tension waveform, features with low correlation to the likelihood of chromophores forming are removed from the normal tension waveform. Similarly, by subtracting the average value from the input tension waveform, features with low correlation to the likelihood of chromophores forming are removed from the input tension waveform. As a result, the accuracy of estimating the likelihood of chromophores forming is improved.

[0031] In one example of the present disclosure, the first pre-processing includes dividing the normal tension waveform into fixed intervals, and the second pre-processing includes dividing the input tension waveform into fixed intervals.

[0032] By dividing the normal tension waveform into fixed sections, it is possible to capture waveforms that are correlated with the likelihood of chromogenic spots occurring. In other words, feature quantities that have a low correlation with the likelihood of chromogenic spots occurring are removed from the normal tension waveform. Similarly, by dividing the input tension waveform into fixed sections, it is possible to capture waveforms that are correlated with the likelihood of chromogenic spots occurring. In other words, feature quantities that have a low correlation with the likelihood of chromogenic spots occurring are removed from the input tension waveform. As a result, the accuracy of estimating the likelihood of chromogenic spots occurring is improved.

[0033] In one example of the present disclosure, the certain section corresponds to a length of the drawn yarn of 6.0 m or less.

[0034] This allows a waveform that is more correlated with the likelihood of mottling to be captured, further improving the accuracy of estimating the likelihood of mottling.

[0035] Another example of the present disclosure provides an estimation method capable of estimating the possibility of dye mottles occurring in a drawn yarn produced by a spinning and drawing apparatus that draws the yarn while heating it. The estimation method includes the steps of acquiring an autoencoder trained to compress a normal tension waveform that shows a normal progression of tension of a drawn yarn different from the drawn yarn and then restore the normal tension waveform, acquiring an input tension waveform that shows a progression of tension of the drawn yarn produced by the spinning and drawing apparatus, and estimating the possibility of dye mottles occurring in the drawn yarn produced by the spinning and drawing apparatus based on the similarity between the input tension waveform and an output tension waveform obtained from the autoencoder by inputting the input tension waveform into the autoencoder.

[0036] This estimation method can estimate the possibility of dye spotting occurring in a drawn yarn produced in a spinning drawing apparatus from an input tension waveform that shows the transition of the tension of the drawn yarn. This makes it possible to estimate the likelihood of dye spotting occurring without conducting a dye test on the produced drawn yarn. As a result, the number of testing steps required to evaluate the quality of the drawn yarn is significantly reduced.

[0037] In one example of the present disclosure, the estimation method further comprises a learning step for generating the autoencoder using the normal tension waveform.

[0038] This allows the learning process and the estimation process to be performed in the same information processing device, i.e., the information processing device can generate an autoencoder within the device itself.

[0039] In one example of the present disclosure, the learning step includes a step of performing a first preprocessing on the normal tension waveform and a step of generating the autoencoder from the normal tension waveform after the first preprocessing, and the estimating step includes a step of performing a second preprocessing on the input tension waveform and a process of inputting the input tension waveform after the second preprocessing into the autoencoder.

[0040] As a result, the normal tension waveform is processed into a format suitable for learning by the first pre-processing, and the input tension waveform is processed into a format suitable for estimation by the second pre-processing.

[0041] In one example of the present disclosure, the first pre-processing includes subtracting the mean value of the normal tension waveform from the normal tension waveform, and the second pre-processing includes subtracting the mean value of the input tension waveform from the input tension waveform.

[0042] The magnitude of tension does not correlate well with the likelihood of chromophores forming. Therefore, by subtracting the average value from the normal tension waveform, features with low correlation to the likelihood of chromophores forming are removed from the normal tension waveform. Similarly, by subtracting the average value from the input tension waveform, features with low correlation to the likelihood of chromophores forming are removed from the input tension waveform. As a result, the accuracy of estimating the likelihood of chromophores forming is improved.

[0043] In one example of the present disclosure, the first pre-processing includes dividing the normal tension waveform into fixed intervals, and the second pre-processing includes dividing the input tension waveform into fixed intervals.

[0044] By dividing the normal tension waveform into fixed sections, it is possible to capture waveforms that are correlated with the likelihood of chromogenic spots occurring. In other words, feature quantities that have a low correlation with the likelihood of chromogenic spots occurring are removed from the normal tension waveform. Similarly, by dividing the input tension waveform into fixed sections, it is possible to capture waveforms that are correlated with the likelihood of chromogenic spots occurring. In other words, feature quantities that have a low correlation with the likelihood of chromogenic spots occurring are removed from the input tension waveform. As a result, the accuracy of estimating the likelihood of chromogenic spots occurring is improved.

[0045] In one example of the present disclosure, the certain section corresponds to a length of the drawn yarn of 6.0 m or less.

[0046] This allows a waveform that is more correlated with the likelihood of mottling to be captured, further improving the accuracy of estimating the likelihood of mottling.

[0047] The above and other objects, features, aspects and advantages of the present invention will become apparent from the following detailed description of the invention taken in conjunction with the accompanying drawings.

[0048] 1 is a schematic diagram showing an example of the device configuration of a yarn take-up machine. FIG. 2 is a schematic diagram showing a measurement device that performs tension measurements to determine the dyeing characteristics of a drawn yarn. FIG. 3 is a block diagram showing the electrical configuration of the measurement device shown in FIG. 2. FIG. 4 is a graph showing tension measurement results for three types of drawn yarns. FIG. 5 is a dyed image of samples created using three types of drawn yarns. FIG. 6 is a diagram showing an example of a process for estimating the likelihood of dye spotting from the tension waveform of a drawn yarn. FIG. 7 is a diagram showing an example of an input tension waveform and an output tension waveform. FIG. 8 is a diagram showing other examples of an input tension waveform and an output tension waveform. FIG. 9 is a diagram showing an example of the hardware configuration of a control unit. FIG. 10 is a diagram showing an example of a learning dataset. FIG. 11 is a diagram showing an example of the functional configuration of a control unit. FIG. 12 is a diagram conceptually showing an example of preprocessing by a preprocessor. FIG. 13 is a diagram conceptually showing another example of preprocessing by a preprocessor. FIG. 14 is a diagram conceptually showing learning processing by a learning unit. FIG. 15 is a diagram conceptually showing estimation processing by an estimation unit. FIG. 16 is a diagram showing the relationship between the similarity between an input tension waveform and an output tension waveform and a threshold. FIG. 17 is a flowchart showing the flow of the learning processing. FIG. 18 is a flowchart showing the flow of the estimation processing. FIG. 19 is a graph showing the results of an evaluation experiment. FIG. 20 is a graph showing the results of an evaluation experiment. It is a figure which shows the result of the evaluation experiment in a graph. It is a figure which shows an example of the device configuration of the information processing system according to the 1st modification. It is a figure which shows the spun yarn take-up machine according to the 2nd modification. It is a figure which shows the spun yarn take-up machine according to the 3rd modification.

[0049] 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 description thereof will not be repeated. Note that each embodiment and each modified example described below may be selectively combined as appropriate.

[0050] <A. Yarn take-up machine 1> First, a yarn take-up machine 1, which is a drawn yarn manufacturing device, will be described with reference to Fig. 1. Fig. 1 is a schematic diagram showing an example of the device configuration of the yarn take-up machine 1. The direction perpendicular to the plane of Fig. 1 is the front-rear direction, and the left-right direction on the plane of Fig. 1 is the left-right direction. The direction perpendicular to both the front-rear direction and the left-right direction is the up-down direction in which gravity acts.

[0051] As shown in Figure 1, the yarn take-off machine 1 mainly comprises a spinning drawing device 3 and a yarn winding device 4. The spinning draw-off machine 1 is configured such that multiple yarns Y spun from a spinning device 2 are drawn by the spinning drawing device 3 to produce drawn yarns Yd, and then the drawn yarns Yd are wound by the yarn winding device 4 to form multiple packages P. The spinning device 2 produces multiple yarns Y by continuously spinning a molten fiber material such as polyester. The multiple yarns Y spun from the spinning device 2 are fed to the spinning drawing device 3 via guide rollers 11. The multiple yarns Y (drawn yarns Yd) drawn by the spinning drawing device 3 are fed to the yarn winding device 4 by guide rollers 12 to 14.

[0052] The spinning and drawing device 3 is a device that heats and draws a plurality of yarns Y, and is disposed below the spinning device 2. The spinning and drawing device 3 has a plurality of godet rollers 31 to 35 housed inside an insulating box 20.

[0053] Each godet roller 31 to 35 is a roller that heats the yarn Y and feeds the yarn Y in the yarn running direction from the yarn inlet 20a to the yarn outlet 20b of the thermal insulation box 20. The godet rollers 31 to 35 are arranged in this order from the upstream side in the yarn running direction. Multiple yarns Y are wound around each godet roller 31 to 35 at a winding angle of less than 360 degrees. The godet rollers 31 to 35 are each driven to rotate at a predetermined yarn feeding speed by a motor (not shown). Each godet roller 31 to 35 also has a built-in roller heater (not shown) that heats the roller surface.

[0054] Of the godet rollers 31 to 35, the three godet rollers 31 to 33 located upstream in the yarn running direction are preheating rollers that heat the yarn Y before drawing. The roller surface temperatures of the godet rollers 31 to 33 are set to a temperature (e.g., approximately 80 to 100°C) equal to or higher than the glass transition point of the yarn Y. On the other hand, of the godet rollers 31 to 35, the two godet rollers 34 and 35 located downstream in the yarn running direction are heat setting rollers that heat set the drawn yarns Y. The roller surface temperatures of the godet rollers 34 and 35 are set to a temperature (e.g., approximately 130 to 150°C) higher than the roller surface temperatures of the godet rollers 31 to 33. That is, the godet rollers 34 and 35 heat the yarn Y at a higher temperature than the godet rollers 31 to 33. The godet rollers 34 and 35 also feed the yarn at a yarn feeding speed faster than the yarn feeding speed of the godet rollers 31 to 33.

[0055] The plurality of yarns Y introduced into the thermal insulation box 20 through the yarn inlet 20a are first preheated to a temperature at which they can be drawn while being fed by the godet rollers 31 to 33. The preheated plurality of yarns Y are drawn between the godet roller 33 and the godet roller 34. The godet roller 33 is the godet roller located most downstream in the yarn running direction among the three godet rollers 31 to 33 that serve as preheating rollers. The godet roller 34 is the godet roller located most upstream in the yarn running direction among the two godet rollers 34, 35 that serve as heat setting rollers. Thereafter, the plurality of yarns Y are further heated to a high temperature while being fed by the godet rollers 34, 35, and the drawn state is heat set. The plurality of yarns Y (drawn yarns Yd) drawn in this manner are led out of the thermal insulation box 20 through the yarn lead-out port 20b.

[0056] The yarn winding device 4 is a device that winds a plurality of yarns Y (drawn yarns Yd) and is disposed below the spinning and drawing device 3. The yarn winding device 4 mainly includes a bobbin holder 41, a plurality of traverse guides 42, a plurality of fulcrum guides 43, and a contact roller 44.

[0057] The bobbin holder 41 has a cylindrical shape extending in the front-rear direction and is rotationally driven by a motor (not shown). A plurality of bobbins B are attached to the bobbin holder 41 and arranged in the axial direction. A plurality of traverse guides 42 and a plurality of fulcrum guides 43 are provided corresponding to the plurality of bobbins B, respectively. Each traverse guide 42 is provided with a driving force from the motor (not shown) to reciprocate in the front-rear direction, thereby traversing the yarn Y hung on the corresponding fulcrum guide 43.

[0058] The yarn winding device 4 rotates the bobbin holder 41 to simultaneously wind a plurality of yarns Y onto a plurality of bobbins B to produce a plurality of packages P. The contact roller 44 comes into contact with the surfaces of the plurality of packages P to apply a predetermined contact pressure to the surfaces of the packages P, thereby adjusting the shape of the packages P.

[0059] <B. Measuring Device 5> Next, the configuration of the measuring device 5 that performs tension measurements to determine the dye unevenness characteristics of the drawn yarn Yd produced as described above will be described with reference to Figures 2 and 3. The direction perpendicular to the plane of the paper in Figure 2 is the front-to-back direction, and the left-to-right direction on the plane of the paper is the left-to-right direction. The direction perpendicular to both the front-to-back direction and the left-to-right direction is the up-to-down direction in which gravity acts.

[0060] The measuring device 5 is for measuring the tension of the drawn yarn Yd that runs in a heated and stretched state. As shown in Figures 2 and 3, the measuring device 5 includes a measuring unit 6 that measures the tension, an output unit 7 that outputs the measurement results, and a control unit 8 that controls the operation of the measuring device 5 as a whole.

[0061] 2, the measurement unit 6 mainly includes a base 61, a first yarn feed roller pair 62 attached to a front surface 61a of the base 61, a tension sensor 63, a heater 64, and a second yarn feed roller pair 65. The first yarn feed roller pair 62 and the second yarn feed roller pair 65 are capable of winding the drawn yarn Yd. Each roller constituting the first yarn feed roller pair 62 corresponds to the "first roller" in the present invention. Each roller constituting the second yarn feed roller pair 65 corresponds to the "second roller" in the present invention.

[0062] The first yarn feed roller pair 62, the tension sensor 63, the heater 64, and the second yarn feed roller pair 65 are arranged in this order from the upstream side to the downstream side in the running direction of the drawn yarn Yd (hereinafter simply referred to as the "yarn running direction"). In other words, the tension sensor 63 and the heater 64 are arranged between the first yarn feed roller pair 62 and the second yarn feed roller pair 65.

[0063] The first yarn feed roller pair 62 is disposed at the upper end of the front surface 61a of the base 61. The first yarn feed roller pair 62 is rotated by a driving force from a motor 62a (see FIG. 3 ), thereby feeding the drawn yarn Yd in a direction from left to right. The operation of the motor 62a is controlled by the control unit 8. That is, the yarn feed speed of the first yarn feed roller pair 62 is controlled by the control unit 8.

[0064] The tension sensor 63 has a measurement roller 63a and a load cell 63b (see FIG. 3). The measurement roller 63a is located to the right of the first yarn feed roller pair 62. That is, the measurement roller 63a is disposed downstream of the first yarn feed roller pair 62 in the yarn running direction. The drawn yarn Yd fed from left to right by the first yarn feed roller pair 62 is wound around the measurement roller 63a. Downstream of the measurement roller 63a in the yarn running direction, the drawn yarn Yd runs from above to below. The load cell 63b detects the force acting from the drawn yarn Yd on the measurement roller 63a, thereby measuring the tension of the drawn yarn Yd stretched between the first yarn feed roller pair 62 and the second yarn feed roller pair 65. The measured value of the tension of the drawn yarn Yd measured by the load cell 63b is input to the control unit 8.

[0065] The heater 64 is disposed below the measuring roller 63a. The heater 64 extends in the vertical direction. The length of the heater 64 in the vertical direction is, for example, 1000 mm. The drawn yarn Yd traveling from above to below downstream of the measuring roller 63a in the yarn traveling direction passes through the heater 64. The heater 64 heats the drawn yarn Yd stretched between the first yarn feed roller pair 62 and the second yarn feed roller pair 65. The heating temperature of the heater 64 is controlled by the control unit 8.

[0066] The second yarn feed roller pair 65 is disposed below the heater 64. The length along the yarn path between the first yarn feed roller pair 62 and the second yarn feed roller pair 65 is, for example, 1,250 mm. The second yarn feed roller pair 65 is rotated by a driving force from a motor 65a (see FIG. 3 ), and feeds the drawn yarn Yd, which has been fed from above downward, diagonally upward to the left. The operation of the motor 65a is controlled by the control unit 8. That is, the yarn feed speed by the second yarn feed roller pair 65 is controlled by the control unit 8.

[0067] 3, the output unit 7 includes a display 70. The display 70 displays the results of the tension measurements performed by the measurement unit 6 under the control of the control unit 8.

[0068] Next, the control performed by the control unit 8 when the measurement device 5 performs tension measurement for estimating dye spot characteristics will be described.

[0069] The control unit 8 controls the yarn feeding speed by the first yarn feeding roller pair 62 and the yarn feeding speed by the second yarn feeding roller pair 65, and controls the tension state of the drawn yarn Yd between the first yarn feeding roller pair 62 and the second yarn feeding roller pair 65. In this embodiment, the control unit 8 controls the yarn feeding speed by the second yarn feeding roller pair 65 so that it is faster than the yarn feeding speed by the first yarn feeding roller pair 62. Note that the control unit 8 may also control the yarn feeding speed by the second yarn feeding roller pair 65 so that it is the same as the yarn feeding speed by the first yarn feeding roller pair 62. In this embodiment, the control unit 8 causes the drawn yarn Yd to be stretched between the first yarn feeding roller pair 62 and the second yarn feeding roller pair 65. More specifically, the control unit 8 controls the deformation amount of the drawn yarn Yd stretched between the first yarn feeding roller pair 62 and the second yarn feeding roller pair 65 to be within the elastic range. That is, the first yarn feed roller pair 62 and the second yarn feed roller pair 65 can support the drawn yarn Yd in a stretched state with a deformation amount within the elastic range.

[0070] The control unit 8 controls the heating temperature of the heater 64 to be 120°C to 140°C. More specifically, the control unit 8 controls the temperature of the drawn yarn Yd heated by the heater 64 to be 120°C to 140°C. The heating temperature of the heater 64 is set to a temperature lower than the heat setting temperature (the roller surface temperature of the godet rollers 34 and 35, which are heat setting rollers (approximately 130°C to 150°C)) used when producing the drawn yarn Yd. This prevents the molecular structure of the drawn yarn Yd from being affected by heating by the heater 64. Furthermore, polyester yarn Y is usually dyed under high temperature and high pressure at approximately 130°C to 135°C. Therefore, by setting the heating temperature of the heater 64 to be 120°C to 140°C, the tension of the drawn yarn Yd at the temperature during dyeing can be measured.

[0071] That is, in the measuring device 5, the tension sensor 63 can measure the tension of the drawn yarn Yd in a state where it is heated at 120°C to 140°C and stretched so that the amount of deformation is within the elastic range. Also, in the measuring device 5, the tension sensor 63 measures the tension of the drawn yarn Yd in a running state. That is, the tension sensor 63 can measure the tension at multiple points in the longitudinal direction of the drawn yarn Yd.

[0072] In this dyeing unevenness estimation method, the tension of the drawn yarn Yd is measured by the measuring device 5, and the possibility of dyeing unevenness occurring in the drawn yarn Yd is estimated based on the tension of the drawn yarn Yd measured by the measuring device 5. Details of this estimation process will be described later.

[0073] <C. Explanation of the Principle> Next, the principle by which the possibility of dye spotting occurring can be estimated from the measured value of the tension of the drawn yarn Yd will be explained with reference to FIGS. 4 and 5. FIG.

[0074] Fig. 4 shows the results of actually measuring the tension of three types of drawn yarns Yd1 to Yd3 that were running in a stretched state so that the deformation amount was within the elastic range while being heated at 120°C to 140°C. Fig. 5 shows the results of dyeing samples S1 to S3 that were made by knitting the drawn yarns Yd1 to Yd3, respectively.

[0075] More specifically, graph G1 shown in Fig. 4 shows the tension waveform measured for drawn yarn Yd1. Fig. 5(a) shows a dyed image of sample S1 made from drawn yarn Yd1.

[0076] Graph G2 in Fig. 4 shows the tension waveform measured for the drawn yarn Yd2. Fig. 5(b) shows a dyed image of sample S2 made from the drawn yarn Yd2.

[0077] Furthermore, graph G3 shown in Fig. 4 shows the tension waveform measured for drawn yarn Yd3. Fig. 5(c) shows a dyed image of sample S3 made from drawn yarn Yd3.

[0078] No staining spots are observed in the stained image of sample S1 shown in Fig. 5(a). On the other hand, mild staining spots are observed in the stained image of sample S2 shown in Fig. 5(b) compared to the stained image of sample S1. Furthermore, severe staining spots are observed in the stained image of sample S3 shown in Fig. 5(c) compared to the stained image of sample S1 (for example, the area surrounded by the dashed line in Fig. 5(c)).

[0079] Whether or not such dye spots will occur is reflected in the tension waveform shown in Figure 4. More specifically, the tension waveforms shown in graphs G2 and G3 vary more than the tension waveform of graph G1. In particular, graph G3 has some points where the measured tension values ​​vary more than graph G1 (for example, the points surrounded by dashed lines in Figure 4). In other words, it can be seen that the more different the tension waveform is from the normal tension waveform of graph G1, the higher the possibility of dye spots occurring. Focusing on this point, the control unit 8 according to this embodiment estimates the likelihood of dye spots occurring in the drawn yarn based on the tension waveform of the drawn yarn.

[0080] <D. Overview of Estimation Processing> Next, a process for estimating the likelihood of dye spotting from the tension waveform of a drawn yarn will be described with reference to Figures 6 to 8. Figure 6 is a diagram showing an outline of the estimation processing.

[0081] The process of estimating the likelihood of dye spots occurring is executed, for example, by the above-mentioned control unit 8, which is an example of the information processing device 100. More specifically, the control unit 8 acquires the autoencoder 124 generated by a learning process. The autoencoder 124 has been trained in advance to compress a normal tension waveform that shows the normal transition of tension in a drawn yarn and then restore the normal tension waveform. Details of the learning process will be described later.

[0082] Next, the control unit 8 acquires an input tension waveform that indicates the transition of the tension of the drawn yarn Yd produced by the spinning and drawing device 3. The input tension waveform to be estimated is acquired, for example, from the tension sensor 63 (see FIG. 2 ).

[0083] Next, the control unit 8 inputs the acquired input tension waveform to the autoencoder 124. As a result, when a normal tension waveform is input, the autoencoder 124 outputs a tension waveform similar to the normal tension waveform. On the other hand, when an abnormal tension waveform is input, the autoencoder 124 outputs a tension waveform different from the abnormal tension waveform.

[0084] Thereafter, the control unit 8 calculates the similarity between the input tension waveform and the output tension waveform obtained from the autoencoder 124 by inputting the input tension waveform to the autoencoder 124. Then, based on the calculated similarity, the control unit 8 estimates the possibility of dye spots occurring in the drawn yarn Yd that is the estimation target.

[0085] 7 is a diagram showing an example of an input tension waveform and an output tension waveform. In this example, the input tension waveform and the output tension waveform are similar to each other. Thus, the more similar the input tension waveform and the output tension waveform are, the lower the possibility of dye spots occurring.

[0086] 8 is a diagram showing another example of input and output tension waveforms. In this example, the input and output tension waveforms are dissimilar to each other. Thus, the more different the input and output tension waveforms are, the greater the likelihood of dye spots occurring.

[0087] In this way, the degree of similarity between the input tension waveform and the output tension waveform correlates with the likelihood of dye spots occurring in the drawn yarn Yd. Focusing on this point, the control unit 8 estimates the possibility of dye spots occurring from the degree of similarity. This allows the possibility of dye spots occurring to be estimated without actually dyeing the drawn yarn Yd produced in the spinning and drawing device 3. As a result, the number of steps required for testing to evaluate the quality of the drawn yarn Yd can be significantly reduced.

[0088] Furthermore, the autoencoder 124 is generated by a learning process using normal tension waveforms. That is, there is no need to collect abnormal tension waveforms to generate the autoencoder 124, and learning data can be easily collected. This allows for a significant reduction in design man-hours.

[0089] <E. Hardware Configuration of Control Unit 8> Next, the hardware configuration of the control unit 8 shown in Fig. 3 will be described with reference to Fig. 9. Fig. 9 is a diagram showing an example of the hardware configuration of the control unit 8.

[0090] The control unit 8 includes a processor 101, a read only memory (ROM) 102, a random access memory (RAM) 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.

[0091] The processor 101 is configured, for example, by at least one integrated circuit, which may be configured, for example, by 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.

[0092] The processor 101 controls the operation of the control unit 8 by executing various programs such as a learning program 126 and an estimation program 128. Upon receiving an execution command for one of the programs, the processor 101 reads the program to be executed from the auxiliary storage device 120 or the ROM 102 into the RAM 103. The RAM 103 functions as a working memory and temporarily stores various data required for executing the program.

[0093] A LAN (Local Area Network), an antenna, etc. are connected to the communication interface 104. The control unit 8 exchanges data with external devices via the communication interface 104. The external devices include, for example, servers.

[0094] The display 70 (see FIG. 3 ) described above is connected to the display interface 105. The display interface 105 sends image signals for displaying images to the display 70 in accordance with commands from the processor 101 or the like. The display 70 is, for example, a liquid crystal display, an organic EL (Electro Luminescence) display, or other displays. The display 70 may be configured integrally with the control unit 8 or may be configured separately from the control unit 8.

[0095] 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 any other device capable of accepting user operations. The input device 108 may be configured integrally with the control unit 8 or may be configured separately from the control unit 8.

[0096] The auxiliary storage device 120 is a storage medium such as a hard disk, a flash memory, or an SSD (Solid State Drive). The auxiliary storage device 120 stores, for example, a training dataset 122, the above-described autoencoder 124, a training program 126, and an estimation program 128. These may be stored in a storage location other than the auxiliary storage device 120, such as a storage area of ​​the processor 101 (e.g., cache memory), the ROM 102, the RAM 103, an external device (e.g., a server), or the like.

[0097] The learning program 126 is a program for generating the autoencoder 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 a program that does not include some of these modules does not deviate from the spirit of the learning program 126 according to this embodiment. Furthermore, some or all of the functions provided by the learning program 126 may be realized by dedicated hardware. Furthermore, the control unit 8 may be configured in the form of a so-called cloud service in which at least one server executes part of the processing of the learning program 126.

[0098] The estimation program 128 is a program for estimating the likelihood of dye unevenness occurring in a drawn yarn produced by the spinning and drawing apparatus 3 using the autoencoder 124. The estimation program 128 may be provided not as a standalone program but as part of an arbitrary program. In this case, the estimation process by the estimation program 128 is realized in cooperation with the arbitrary program. Even a program that does not include some of these modules does not deviate from the spirit of the estimation program 128 according to this embodiment. Furthermore, some or all of the functions provided by the estimation program 128 may be realized by dedicated hardware. Furthermore, the control unit 8 may be configured in the form of a so-called cloud service in which at least one server executes part of the processing of the estimation program 128.

[0099] <F. Training Data Set 122> Next, the training data set 122 shown in Fig. 9 will be described with reference to Fig. 10. Fig. 10 is a diagram showing an example of the training data set 122.

[0100] The training data set 122 includes a plurality of training data 123. The number of training data 123 included in the training data set 122 is arbitrary. As an example, the number of training data 123 is several tens to several millions.

[0101] The training data set 122 is composed only of training data 123 of normal tension waveforms. More specifically, each training data 123 is associated with a data ID (Identification) and a normal tension waveform in which dysplasia is unlikely to occur. The data ID is information for uniquely identifying the training data 123. The data ID is entered by the user, for example, so that no duplicate data IDs are created.

[0102] The tension waveforms defined in the learning data 123 are a data group in which measured tensions are arranged in chronological order. That is, in one piece of learning data 123, tensions are associated with each other over time.

[0103] The tension of the drawn yarn Yd is measured in a state where the drawn yarn Yd is not slackened and is not stretched while being heated at 120°C to 140°C, or in a state where the drawn yarn Yd is stretched to a deformation amount within the elastic range while being heated at 120°C to 140°C.

[0104] The training data 123 may be collected from various devices. As an example, the training data 123 may include a tension waveform measured by the tension sensor 63 in the measuring device 5 described above. Alternatively, the training data 123 may include a tension waveform measured by a tension sensor 51 (see FIG. 23 or 24 ) described below.

[0105] <G. Functional Configuration of Control Unit 8> Next, the functional configuration of the control unit 8 will be described with reference to Fig. 11 to Fig. 16. Fig. 11 is a diagram showing an example of the functional configuration of the control unit 8.

[0106] 11, the control unit 8 includes, as functional components, a preprocessing unit 151, a learning unit 152, a preprocessing unit 153, and an estimation unit 154. These functional components will be described below in order.

[0107] (G1. Preprocessing Unit 151) First, the function of the preprocessing unit 151 shown in Fig. 11 will be described with reference to Fig. 12 and Fig. 13. Fig. 12 is a diagram conceptually showing an example of preprocessing by the preprocessing unit 151. Fig. 13 is a diagram conceptually showing another example of preprocessing by the preprocessing unit 151.

[0108] The preprocessing unit 151 performs various preprocessing steps (first preprocessing steps) on the normal tension waveform defined in the learning data 123 (see FIG. 10 ) described above, thereby converting the normal tension waveform into a format suitable for learning.

[0109] In one aspect, the preprocessing unit 151 executes the preprocessing shown in FIG. 12 . More specifically, the preprocessing unit 151 calculates the average value of the normal tension waveform. Then, the preprocessing unit 151 subtracts the calculated average value from the normal tension waveform. As explained above in "C. Principle Description," the variation in the tension waveform correlates with the likelihood of chromophores forming. On the other hand, the magnitude of the tension does not correlate as much with the likelihood of chromophores forming. Therefore, by subtracting the average value from the normal tension waveform, only the feature values ​​that are more correlated with the likelihood of chromophores forming remain. As a result, the likelihood of chromophores forming is more accurately estimated.

[0110] In another aspect, the preprocessing unit 151 performs the preprocessing shown in FIG. 13 . More specifically, the preprocessing unit 151 performs preprocessing to divide the normal tension waveform defined in the learning data 123 into fixed intervals ΔL. As an example, the preprocessing unit 151 sequentially slides a frame W of the intervals ΔL on the normal tension waveform, and extracts the tension waveform within the frame W. The effect of this will be explained in "J. Evaluation Experiment" below. The intervals ΔL may or may not overlap each other.

[0111] The length of the section ΔL can be set arbitrarily. Preferably, the section ΔL corresponds to a length of the drawn yarn Yd of 6.0 m or less. The number of tension data points included in this section ΔL is, for example, 200. The effect of this will also be explained in "J. Evaluation Experiment" below.

[0112] It should be noted that the preprocessing by the preprocessing unit 151 does not necessarily have to be performed. Furthermore, the preprocessing unit 151 may perform both the preprocessing shown in Fig. 12 and the preprocessing shown in Fig. 13, or may perform only one of the preprocessing.

[0113] Furthermore, the preprocessing unit 151 may perform preprocessing different from the preprocessing shown in Figures 12 and 13. As an example, the preprocessing unit 151 may perform trend removal processing on the normal tension waveform defined in the learning data 123. Examples of trend removal processing include taking the logarithm of each tension in the time series included in the normal tension waveform, taking the square root of each tension, and subtracting a moving average from each tension.

[0114] (G2. Learning Unit 152) Next, the function of the learning unit 152 shown in Fig. 11 will be described with reference to Fig. 14. Fig. 14 is a diagram conceptually showing the learning process performed by the learning unit 152.

[0115] The learning unit 152 executes a learning process to generate the autoencoder 124 from a normal tension waveform. The normal tension waveform is, for example, a waveform that has been preprocessed by the preprocessing unit 151. The machine learning algorithm used in the learning process is not particularly limited, and may be, for example, a neural network such as deep learning. The learning process using a neural network will be described below.

[0116] As shown in FIG. 14, the autoencoder 124 is composed of an input layer X, an intermediate layer H, and an output layer Y.

[0117] The input layer X is configured to receive the normal tension waveform after preprocessing by the preprocessing unit 151. The input layer X includes, for example, N units x １ ~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 will be 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 intermediate layer H.

[0118] The intermediate layer H is composed of one layer or multiple layers. In the example of FIG. 14, the intermediate layer H is composed of L layers (L is a natural number). Each layer of the intermediate layer H includes multiple units. The number of units in each layer of the intermediate layer H may be the same or different. In the example of FIG. 14, the first layer of the intermediate 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).

[0119] 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 (or from) the accumulation result, inputs the addition result (or subtraction result) to a predetermined function (for example, a sigmoid function), and outputs the output value of the function to each unit in the next layer.

[0120] In the autoencoder 124, 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 transmission from the input layer X to the hidden layer H.

[0121] The output layer Y is configured to restore the tension waveform compressed in the intermediate layer H. More specifically, the output layer Y is configured with the same number of units as the input layer X. As an example, if the input layer X is configured with N units, the output layer Y is also configured with N units. In the example of FIG. 14, the output layer Y is configured with N units y １ ~y Ｎ In the following, the unit y １ ~y Ｎ is also referred to as unit y.

[0122] Each of the units y is connected to each unit h in the final layer of the intermediate layer H. Ｌ１ ~h ＬＲ Each of the units y receives an output value from each unit in the final layer of the intermediate layer H, multiplies each output value by a weight, accumulates the results of these multiplications, adds (or subtracts) a predetermined bias to (or from) the accumulated result, inputs the result of the addition (or subtraction) to a predetermined function (for example, a Sigmonite function), and outputs the output result of the function as an output value.

[0123] Next, the update process of the internal parameters of the autoencoder 124 by the learning unit 152 will be described.

[0124] The learning unit 152 inputs the first normal tension waveform T(t) to the autoencoder 124. As a result, the autoencoder 124 compresses the tension waveform T(t) and restores it to a tension waveform T'(t) of the same dimension. Next, the learning unit 152 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):

[0125] Z = {(T(t １ )-T'(t １ )) ２ +...+(T(t Ｎ )-T'(t Ｎ )) ２} / N (1) Next, the learning unit 152 updates the internal parameters (e.g., weights and biases) of the autoencoder 124 so as to reduce the error "Z". The updating of the internal parameters is realized by, for example, backpropagation.

[0126] The learning unit 152 repeatedly updates the internal parameters of the autoencoder 124 for each normal tension waveform to be learned. As a result, the autoencoder 124 learns to compress a normal tension waveform and then restore the normal tension waveform. In other words, when a normal tension waveform is input, the autoencoder 124 outputs a tension waveform similar to the normal tension waveform, and when an abnormal tension waveform is input, the autoencoder 124 outputs a tension waveform different from the abnormal tension waveform. In other words, the autoencoder 124 functions like a kind of filter that passes normal tension waveforms but does not pass abnormal tension waveforms.

[0127] (G3. Pre-processing unit 153) Next, the function of the pre-processing unit 153 shown in FIG. 11 will be described.

[0128] The preprocessing unit 153 performs various preprocessing operations (second preprocessing operations) on the input tension waveform acquired from the above-described tension sensor 63. Typically, the preprocessing unit 153 performs the same preprocessing operations as the above-described preprocessing unit 151.

[0129] As an example, the preprocessing unit 153 performs preprocessing by calculating the average value of the input tension waveform and subtracting the calculated average value from the input tension waveform. This preprocessing has been described with reference to FIG. 12 , and therefore will not be described again.

[0130] As another example, the pre-processing unit 153 performs pre-processing to divide the input tension waveform into fixed sections. This processing has been described with reference to FIG. 13 , and therefore, description thereof will not be repeated.

[0131] (G4. Estimation Unit 154) Next, the function of the estimation unit 154 shown in Fig. 11 will be described with reference to Fig. 15 and Fig. 16. Fig. 15 is a diagram conceptually showing the estimation process performed by the estimation unit 154.

[0132] The estimation unit 154 estimates the possibility of dye spots occurring in the drawn yarn produced by the spinning and drawing device 3 based on the similarity between the input tension waveform after preprocessing by the above-mentioned preprocessing unit 153 and the output tension waveform obtained by inputting the input tension waveform to the autoencoder 124.

[0133] More specifically, the estimation unit 154 first acquires the autoencoder 124 and inputs the input tension waveform after preprocessing by the preprocessing unit 153 to the autoencoder 124. The acquisition source of the autoencoder 124 may be a storage device within the control unit 8 or an external device. The autoencoder 124 receives an input tension waveform and outputs an output tension waveform. At this time, if a normal tension waveform is input, the autoencoder 124 outputs a tension waveform similar to the normal tension waveform, and if an abnormal tension waveform is input, the autoencoder 124 outputs a tension waveform different from the abnormal tension waveform.

[0134] The estimation unit 154 then calculates the similarity between the input tension waveform and the output tension waveform. Any algorithm can be used to calculate the similarity. For example, the algorithm for calculating the similarity may be a mean squared error (MSE), a sum of squared differences (SSD), a sum of absolute differences (SAD), normalized cross-correlation (NCC), or a zero-mean normalized cross-correlation (ZNCC).

[0135] The magnitude of the calculated similarity may vary depending on the algorithm used. That is, the calculated similarity value may be greater the more similar the input tension waveform and the output tension waveform are to each other, or may be smaller the more similar the input tension waveform and the output tension waveform are to each other.

[0136] Fig. 16 is a diagram showing the relationship between the similarity between the input tension waveform and the output tension waveform and the threshold value. The similarity shown in Fig. 16 is calculated using mean square error. In mean square error, the more similar the input tension waveform and the output tension waveform are, the smaller the calculated similarity will be. In other words, the more different the input tension waveform and the output tension waveform are, the greater the calculated similarity will be.

[0137] The similarity is calculated for each input tension waveform divided into fixed sections by preprocessing by the preprocessing unit 153. If the number of similarities exceeding a predetermined threshold among the calculated similarities is equal to or greater than a predetermined number, the estimation unit 154 determines that there is a high possibility of dysmetropia occurring and outputs an estimation result indicating abnormality. On the other hand, if the number of similarities exceeding a predetermined threshold among the calculated similarities is less than the predetermined number, the estimation unit 154 determines that there is a low possibility of dysmetropia occurring and outputs an estimation result indicating normality.

[0138] The threshold value may be set in advance or may be arbitrarily set by the user. The predetermined number may be set in advance or may be arbitrarily set by the user. As an example, the predetermined number is set to "1."

[0139] The estimation result by the estimation unit 154 may be output in any manner. As one example, the estimation result may be displayed as a message on the display 70 described above. As another example, the estimation result may be output as audio. As yet another example, the estimation result may be stored as a log in a storage device within the control unit 8.

[0140] The granularity of the estimation result by the estimation unit 154 is arbitrary. As one example, the estimation result is output as two values: a normal estimation result indicating a low possibility of dysmetropia occurring, and an abnormal estimation result indicating a high possibility of dysmetropia occurring. As another example, the estimation result may be output as a numerical value indicating the possibility of dysmetropia occurring.

[0141] <H. Flowchart of Learning Process> Next, a flowchart of the learning process performed by the control unit 8 will be described with reference to Fig. 17. Fig. 17 is a flowchart showing the flow of the learning process.

[0142] The processor 101 of the control unit 8 executes the above-described learning program 126 (see FIG. 9) to perform the various processes shown in FIG. 17. In another aspect, some or all of the processes shown in FIG. 17 may be performed by circuit elements or other hardware.

[0143] In step S110, the processor 101 functions as the above-mentioned preprocessing unit 151 (see FIG. 11 ) and performs predetermined preprocessing on the normal tension waveform defined in each of the learning data 123. Since the preprocessing is as described above, its description will not be repeated.

[0144] In step S112, the processor 101 inputs the normal tension waveform after preprocessing in step S110 to the autoencoder 124 described above.

[0145] In step S114, the processor 101 functions as the learning unit 152 (see FIG. 11 ) described above, calculates the error between the normal tension waveform input to the autoencoder 124 and the output tension waveform output from the autoencoder 124, and updates the internal parameters of the autoencoder 124 so that this error becomes smaller than the current error. The parameters are updated by, for example, the error backpropagation method. The learning process in step S114 is as described above, and therefore will not be described again.

[0146] In step S120, the processor 101 determines whether to end the learning process. As an example, the processor 101 determines to end the learning process when the estimation accuracy using the test data exceeds a desired accuracy. Alternatively, the processor 101 determines to end the learning process when the number of updates of the internal parameters of the autoencoder 124 exceeds a predetermined number.

[0147] If processor 101 determines to end the learning process (YES in step S120), it ends the process shown in Fig. 17. Otherwise (NO in step S120), processor 101 returns control to step S112.

[0148] <I. Flowchart of Estimation Processing> Next, a flowchart of the estimation processing performed by the control unit 8 will be described with reference to Fig. 18. Fig. 18 is a flowchart showing the flow of the estimation processing.

[0149] The processor 101 of the control unit 8 executes the above-described estimation program 128 (see FIG. 9 ) to perform the various processes shown in FIG. 18. In another aspect, some or all of the processes shown in FIG. 18 may be performed by circuit elements or other hardware.

[0150] In step S210, the processor 101 acquires an input tension waveform that indicates the transition of the tension of the drawn yarn Yd that is the estimation target. The input tension waveform is acquired from the tension sensor 63, for example.

[0151] In step S212, the processor 101 functions as the above-mentioned preprocessing unit 153 (see FIG. 11 ) and performs predetermined preprocessing on the input tension waveform acquired in step S210. As one example, the processor 101 performs preprocessing to calculate the average value of the input tension waveform and subtract the calculated average value from the input tension waveform. As another example, the processor 101 performs preprocessing to divide the input tension waveform into sections ΔL (see FIG. 13 ).

[0152] In step S214, the processor 101 inputs the input tension waveform after the preprocessing in step S212 to the autoencoder 124 described above.

[0153] In step S216, the processor 101 functions as the above-mentioned estimation unit 154 (see FIG. 11 ) and calculates the similarity between the input tension waveform input to the autoencoder 124 and the output tension waveform output from the autoencoder 124.

[0154] In step S220, the processor 101 determines whether similarities have been calculated for all input tension waveforms separated by a fixed interval ΔL in the preprocessing in step S212. If the processor 101 determines that similarities have been calculated for all of the separated input tension waveforms (YES in step S220), the processor 101 switches control to step S230. If not (NO in step S230), the processor 101 returns control to step S214.

[0155] In step S230, the processor 101 functions as the estimation unit 154 described above and determines whether at least one of the similarities calculated in step S216 satisfies an abnormality condition. The abnormality condition is satisfied when the input tension waveform and the output tension waveform are not similar. As an example, the abnormality condition is satisfied when one of the calculated similarities exceeds a predetermined threshold. If the processor 101 determines that at least one of the similarities calculated in step S216 satisfies the abnormality condition (YES in step S230), the processor 101 switches control to step S232. If not (NO in step S230), the processor 101 switches control to step S234.

[0156] In step S232, the processor 101 functions as the estimation unit 154 described above, and outputs an estimation result of an abnormality indicating that there is a high possibility of dysmetropia occurring.

[0157] In step S234, the processor 101 functions as the above-described estimation unit 154 and outputs a normal estimation result indicating that the possibility of dysmetastasis is low.

[0158] 13, the preprocessing units 151 and 153 perform preprocessing to divide the tension waveform into fixed intervals ΔL. The inventors have found a preferable value for the intervals ΔL through evaluation experiments.

[0159] The evaluation experiment will be described below with reference to Figures 19 to 23. Figures 19 to 23 show the results of the evaluation experiment in the form of graphs GA to GE.

[0160] (J1. Experimental Conditions) First, the conditions of the experiments conducted by the inventors will be explained.

[0161] The inventors prepared five normal tension waveforms where no dye spots occurred (hereinafter also referred to as "first tension waveforms"), two normal tension waveforms where mild dye spots occurred (hereinafter also referred to as "second tension waveforms"), and two normal tension waveforms where severe dye spots occurred (hereinafter also referred to as "third tension waveforms"). The inventors then conducted evaluation experiments using the following steps (1) to (4).

[0162] (1) Preprocessing is performed to divide each of the first tension waveforms into sections ΔL, and the above-described autoencoder 124 is generated from the preprocessed first tension waveforms.

[0163] (2) Preprocessing is performed to divide each of the first to third tension waveforms into sections ΔL, and the preprocessed first to third tension waveforms are input to the autoencoder 124.

[0164] (3) The similarity between the input tension waveform input to the autoencoder 124 and the output tension waveform output from the autoencoder 124 is calculated, and the similarity is converted into a box plot and plotted on graphs GA to GE.

[0165] (4) The length of the section ΔL is changed and the above steps (1) to (3) are repeated.

[0166] 19 and 20, the length of the section ΔL is shown as the number of steps. The number of steps corresponds to the number of tension data included in the section ΔL. The inventors conducted evaluation experiments by changing the number of steps in the above step (4) from "50," "100," and "200."

[0167] (J2. Graphs GA to GE) Next, the meanings of graphs GA to GE shown in FIGS. 19 to 21 will be explained.

[0168] Graph GA shown in Fig. 19 shows the experimental results when the number of steps was set to "50." The number of steps "50" corresponds to a length of 1.5 m of the drawn yarn Yd.

[0169] Graph GB shown in Fig. 20 shows the experimental results when the number of steps was set to "100." The number of steps "100" corresponds to a length of 3.0 m of the drawn yarn Yd.

[0170] Graph GC shown in Fig. 21 shows the experimental results when the number of steps was set to 200. The number of steps "200" corresponds to a length of 6.0 m of the drawn yarn Yd.

[0171] The vertical axes of graphs GA to GC represent the similarity calculated in step (3) above. "Normal" on the horizontal axes of graphs GA to GC represents the experimental results for the first tension waveform. "Dyeing spot (mild)" on the horizontal axes of graphs GA to GC represents the experimental results for the second tension waveform. "Dyeing spot (severe)" on the horizontal axes of graphs GA to GC represents the experimental results for the third tension waveform.

[0172] The lower limits of the boxes shown in graphs GA to GC correspond to the bottom quarter value (first quartile) when similarities are sorted in ascending order. The dashed lines within the boxes shown in graphs GA to GC correspond to the bottom half value (median) when similarities are sorted in ascending order. The upper limits of the boxes shown in graphs GA to GC correspond to the bottom three-quarter value (third quartile) when similarities are sorted in ascending order.

[0173] Additionally, the upper limit of the whiskers shown in graphs GA to GC corresponds to "first quartile - interquartile range x 1.5." The lower limit of the whiskers shown in graphs GA to GC corresponds to "third quartile + interquartile range x 1.5." The interquartile range corresponds to "third quartile - first quartile." The plotted points shown in graphs GA to GC indicate "outliers" that are not included within the range of the whiskers.

[0174] (J3. Analysis Results) Next, the analysis results of graphs GA to GC will be explained.

[0175] As shown in graphs GA to GC, when the number of steps is 200 or less, the distribution of similarities for "normal" is more concentrated than the distributions of similarities for "dyeing spots (mild)" and "dyeing spots (severe)." Furthermore, the outliers for similarities for "normal" are fewer than the outliers for "dyeing spots (mild)" and "dyeing spots (severe)." In other words, by setting the number of steps to 200 or less, the characteristics of the waveform due to dyeing spots can be more reliably captured, and the likelihood of dyeing spots occurring can be estimated with higher accuracy.

[0176] The number of steps is preferably set to, for example, several to ten times (approximately 1 to 5 m) the length of the staining abnormality (several tens of cm). More preferably, the number of steps is set to 100 or less. Even more preferably, the number of steps is set to 50 or less.

[0177] <K. First Modification> Next, a first modification of the above embodiment will be described with reference to Fig. 22. Fig. 22 is a diagram showing an example of the device configuration of an information processing system 500 in this modification.

[0178] In the following description, the same or similar components as those in the above-described embodiment will be denoted by the same reference numerals in the drawings, and the description thereof may be omitted.

[0179] 11 , the functional configurations of the preprocessing unit 151, the learning unit 152, the preprocessing unit 153, and the estimation unit 154 are implemented in the control unit 8 of the measurement device 5. However, these functional configurations do not necessarily have to be implemented in the control unit 8. As an example, some or all of these functional configurations may be implemented in the server 200, which is an example of the information processing device 100.

[0180] As shown in Fig. 22, an information processing system 500 includes one or more measuring devices 5 and one or more servers 200. In the example of Fig. 22, the information processing system 500 is composed of three measuring devices 5A to 5C and one server 200. Hereinafter, when there is no need to particularly distinguish between the measuring devices 5A to 5C, the measuring devices 5A to 5C will be referred to as measuring devices 5.

[0181] The measurement device 5 and the server 200 are connected to the same network NW and are configured to be able to communicate with each other. The measurement device 5 and the server 200 may be connected for communication via a wired or wireless connection.

[0182] The server 200 is a desktop personal computer, a laptop personal computer, a tablet terminal, or any other information processing terminal. In the example of Fig. 22 , the preprocessing unit 151, the learning unit 152, the preprocessing unit 153, and the estimation unit 154 are implemented in the server 200.

[0183] The server 200 collects the above-mentioned training data 123 (see FIG. 10 ) from the measurement devices 5 (e.g., measurement devices 5A and 5B) connected to the network NW. Next, the training unit 152 of the server 200 executes a training process using the training data 123 that has been preprocessed by the preprocessing unit 151, and generates the above-mentioned autoencoder 124. The training process is as described above, and therefore will not be described again.

[0184] Thereafter, the server 200 receives the input tension waveform measured by the measuring device 5C. Next, the estimation unit 154 of the server 200 inputs the input tension waveform that has been preprocessed by the preprocessing unit 153 to the autoencoder 124. Next, the estimation unit 154 of the server 200 estimates the possibility of dye spotting occurring in the drawn yarn based on the similarity between the input tension waveform and the output tension waveform obtained from the autoencoder 124. Since this estimation process is as described above, its description will not be repeated. Thereafter, the server 200 transmits the estimation result to the measuring device 5C.

[0185] <L. Second Modification> Next, a second modification of the above embodiment will be described with reference to Fig. 23. Fig. 23 is a diagram showing a yarn take-up machine 1a according to the second modification.

[0186] 2, an example in which the tension sensor 63 is provided inside the measurement unit 6 has been described, but the installation position of the tension sensor 63 is not limited to inside the measurement unit 6. For example, as shown in Fig. 23, the tension sensor 51 may be provided inside the yarn take-up machine 1a. Points other than the location of the tension sensor 51 are as described above, and therefore, description thereof will not be repeated below.

[0187] The yarn take-off machine 1a according to this modified example has multiple godet rollers 31 to 35. As described above, the three godet rollers 31 to 33 located upstream in the yarn running direction are preheating rollers that heat the yarn Y before drawing. On the other hand, the two godet rollers 34 and 35 located downstream in the yarn running direction are heat setting rollers that heat set the multiple drawn yarns Y.

[0188] The tension sensor 51 is disposed so as to measure the tension of the drawn yarn Yd traveling between the godet rollers 31 to 33, which are preheating rollers, and the godet roller 35, which is the heat setting roller and is located most downstream in the yarn traveling direction, of the godet rollers 34 and 35. In the example of Fig. 23, the tension sensor 51 is disposed between the godet roller 34 and the godet roller 35, and measures the tension of the drawn yarn Yd in a state where it has been heated by the godet rollers 34 and 35, which are heat setting rollers.

[0189] The yarn feeding speed of the godet rollers 34, 35 is set so that the drawn yarn Yd between the godet rollers 34 and 35 is in a state where it is neither loose nor stretched, or is stretched to a deformation amount within the elastic range.

[0190] By providing the tension sensor 51 inside the yarn take-up machine 1a, the information processing device 100 can acquire an input tension waveform for estimating the possibility of dyeing unevenness occurring during the manufacturing process of the drawn yarn Yd, thereby shortening the time required to estimate the possibility of dyeing unevenness occurring.

[0191] <M. Third Modification> Next, a third modification of the above embodiment will be described with reference to Fig. 24. Fig. 24 is a diagram showing a yarn take-up machine lb according to the third modification.

[0192] In the example of FIG. 23 described above, the tension sensor 51 is provided between the godet roller 34 and the godet roller 35, but the tension sensor 51 may be provided in other locations.

[0193] More specifically, in the yarn take-up machine 1b according to this modified example, a tension sensor 51 and a heater 52 are provided between the guide roller 13 and the guide roller 14. The tension sensor 51 is arranged to measure the tension of the drawn yarn Yd running between the guide roller 13 (first roller) and the guide roller 14 (second roller). The heater 52 is configured to heat the drawn yarn Yd stretched between the guide roller 13 and the guide roller 14.

[0194] The tension sensor 51 measures the tension of the drawn yarn Yd in a state heated by the heater 52. The yarn feeding speed of the guide rollers 13, 14 is set so that the drawn yarn Yd between the guide rollers 13, 14 is in a state where it is neither slackened nor stretched, or is stretched with a deformation amount within the elastic range.

[0195] The embodiments disclosed herein should be considered to be illustrative in all respects and not restrictive. The scope of the present invention is defined by the claims, not by the above description, and is intended to include all modifications within the meaning and scope of the claims.

[0196] 3 Spinning drawing device 5 Measuring device 31 to 33 Godet rollers (preheating rollers) 34, 35 Godet rollers (heat setting rollers) 51 Tension sensor 52 Heater 62 First yarn feed roller pair (first roller) 63 Tension sensor 64 Heater 65 Second yarn feed roller pair (second roller) 100 Information processing device 101 Processor 124 Autoencoder 128 Estimation program Yd Drawn yarn ΔL section

Claims

1. An estimation program capable of estimating the possibility of dye spots occurring in a drawn yarn produced by a spinning and drawing device that draws the yarn while heating it, the estimation program causing an information processing device to execute the steps of: acquiring an autoencoder that has been trained to compress a normal tension waveform that shows the normal tension progression of a drawn yarn different from the drawn yarn, and then restoring the normal tension waveform; acquiring an input tension waveform that shows the tension progression of the drawn yarn produced by the spinning and drawing device; and estimating the possibility of dye spots occurring in the drawn yarn produced by the spinning and drawing device based on the similarity between the input tension waveform and an output tension waveform obtained from the autoencoder by inputting the input tension waveform into the autoencoder.

2. The estimation program according to claim 1, further causing the information processing device to execute a learning step of generating the autoencoder using the normal tension waveform.

3. The estimation program according to claim 2, wherein the learning step includes a step of performing a first preprocessing on the normal tension waveform and a step of generating the autoencoder from the normal tension waveform after the first preprocessing, and the estimating step includes a step of performing a second preprocessing on the input tension waveform and a process of inputting the input tension waveform after the second preprocessing into the autoencoder.

4. The estimation program according to claim 3, wherein the first pre-processing includes a process of subtracting an average value of the normal tension waveform from the normal tension waveform, and the second pre-processing includes a process of subtracting an average value of the input tension waveform from the input tension waveform.

5. The estimation program according to claim 3 or 4, wherein the first pre-processing includes a process of dividing the normal tension waveform into fixed intervals, and the second pre-processing includes a process of dividing the input tension waveform into the fixed intervals.

6. The estimation program according to claim 5, wherein the certain section corresponds to a length of the drawn yarn of 6.0 m or less.

7. An information processing device capable of estimating the possibility of dye spots occurring in a drawn yarn produced by a spinning and drawing device that draws the yarn while heating it, comprising a processor for controlling the information processing device, wherein the processor executes the following processes: acquiring an autoencoder that has been trained to restore a normal tension waveform that shows the normal tension progression of a drawn yarn different from the drawn yarn after compressing the normal tension waveform; acquiring an input tension waveform that shows the tension progression of the drawn yarn produced by the spinning and drawing device; and estimating the possibility of dye spots occurring in the drawn yarn produced by the spinning and drawing device based on the similarity between the input tension waveform and an output tension waveform obtained from the autoencoder by inputting the input tension waveform into the autoencoder.

8. An information processing device as described in claim 7, wherein the input tension waveform is measured by a measuring device different from the spinning and drawing device, the measuring device comprising: first and second rollers around which a yarn can be passed; a heater disposed between the first roller and the second roller for heating the yarn passed between the first roller and the second roller; and a tension sensor disposed to measure the tension of the drawn yarn running between the first roller and the second roller.

9. The information processing device according to claim 7, wherein the spinning and drawing device further comprises: a pre-heating roller for heating the yarn before drawing; at least one heat setting roller arranged downstream of the pre-heating roller in the yarn running direction, heating the yarn to a higher temperature than the pre-heating roller and feeding the yarn at a yarn feeding speed faster than the yarn feeding speed by the pre-heating roller; and a tension sensor arranged to measure the tension of the drawn yarn traveling between the pre-heating roller and the heat setting roller located most downstream in the yarn running direction among the at least one heat setting roller.

10. The information processing device according to claim 7, wherein the spinning and drawing device further comprises: first and second rollers around which the yarn can be passed; a heater disposed between the first roller and the second roller for heating the yarn passed between the first roller and the second roller; and a tension sensor disposed to measure the tension of the drawn yarn running between the first roller and the second roller.

11. An estimation method capable of estimating the possibility of dye spots occurring in a drawn yarn produced by a spinning and drawing apparatus in which the yarn is drawn while being heated, comprising the steps of: acquiring an autoencoder that has been trained to restore a normal tension waveform that shows the normal tension progression of a drawn yarn different from the drawn yarn after compressing the normal tension waveform; acquiring an input tension waveform that shows the tension progression of the drawn yarn produced by the spinning and drawing apparatus; and estimating the possibility of dye spots occurring in the drawn yarn produced by the spinning and drawing apparatus based on the similarity between the input tension waveform and an output tension waveform obtained from the autoencoder by inputting the input tension waveform into the autoencoder.

Images