Equipment and methods for detecting defects in roving, and methods for managing spinning production lines and their spinning machines.

Abstract

An apparatus for detecting defects in drafted roving (R) processed in a spinning machine is provided. The apparatus includes a data acquisition device (60) and a processing device (70) for digitally processing the acquired image (I) and detecting defects using a machine learning detection algorithm. The invention also provides a method for detecting defects in drafted roving (R) processed in a spinning machine, a method for managing a spinning machine (1) in a spinning production line, and a method for managing a spinning production line.

Description

Technical Field

[0001] This invention pertains to the field of textile fiber processing, and more particularly to the field of instruments and methods for detecting defects in products during the spinning preparation process. Specifically, the object of this invention is a method and apparatus for detecting entanglements or knots, commonly referred to as "neps," in yarns produced in ring spinning machines. Background Technology

[0002] As is well known, after drafting and twisting the roving, the spinning machine can process the roving bobbin to obtain the yarn spool.

[0003] For this purpose, the spinning machine includes a frame, a drafting device supported by the frame, and a guide rail. The frame extends along a longitudinal axis and supports a warp beam that suspends the roving bobbins. The drafting device includes a set of connecting tubular members with longitudinal extensions. The roving to be drafted passes through the connecting tubular members. The guide rail carries the spindles, which are arranged in rows along the longitudinal axis and rotate about the vertical axis of the spindles. The drafted and twisted yarn is wound on the spindles.

[0004] It is well known how defects in yarn can have an unpleasant impact on the appearance of fabrics, especially in the case of colored fabrics. For this reason, there is a strong need within the industry to monitor the extent and frequency of defects, particularly fiber entanglement.

[0005] To date, there are two industrial methods for detecting defects in yarn.

[0006] The first method involves analyzing samples in a laboratory setting, consisting of yarn from spools removed from a spinning machine. This is typically achieved using capacitive sensors capable of detecting changes in yarn mass along its length to determine the type and frequency of entanglement. While generally reliable, this method does not allow for intervention with processing parameters to improve yarn quality, nor can it be used to understand the cause of any defects found, as it is performed after yarn production is complete. For example, the USTER® Tester 5-S800, manufactured and sold by Uster Technologies AG, is frequently used.

[0007] The second method involves using a detection module based on capacitance or optical sensors, positioned at the spindle used for winding the yarn and equipped with blades capable of physically eliminating tangles from the yarn. Even in this case, it is impossible to trace the cause of the defect and intervene accordingly.

[0008] Other methods are applied to the yarn winding stage, a process that occurs downstream of the spinning machine; for example, the Uster® Quantum 3 tester, based on a capacitive sensor and manufactured and sold by Uster Technologies AG, and the YarnMaster Prisma tester, based on an optical sensor and manufactured and sold by Loepfe Brothers Ltd, are widely used.

[0009] Furthermore, the primary objective of this invention is to detect defects in yarn within a spinning machine in order to modify the processing parameters of the spinning machine or other upstream machines or maintenance procedures, thereby obtaining higher quality yarn.

[0010] The methods described above are not suitable for this purpose, partly because they use capacitors or optical sensors that require a regular yarn supply.

[0011] However, between the drafting and winding units of the spinning machine, the yarn experiences intense vibration due to the winding and twisting action that occurs downstream.

[0012] Some studies also involve using cameras to acquire images and then processing them to detect tangles. For example, some methods are described in the following articles:

[0013] - Li Z, Pan R, and Gao W. Formation of digital yarn blackboard using sequence images. Textile Research Journal. 2016; 86: 593-603;

[0014] - Eldessouki M, Ibrahim S, and Militky J. A dynamic and robust imageprocessing based method for measuring the yarn diameter and its variation. Textile Research Journal. 2014; 84: 1948-60;

[0015] - Digital image processing of cotton yarn seriplane, Ling C, Liangying Z, Li C, and Xuanli Z. 2010 International Conference on Computer and Information Application. 2010, pp. 274-7;

[0016] - Li Z, Xiong N, Wang J, Pan R, Gao W, and Zhang N. An intelligent computer method for automatic mosaic of sequential slub yarn images based on image processing. Textile Research Journal. 2018; 88: 2854-66;

[0017] - Carvalho V, Soares F, and Vasconcelos R, Artificial intelligence and image processing based techniques: A tool for yarns parameterization and fabric prediction. 2009 IEEE Conference on Emerging Technologies & Factory Automation. 2009, pp. 1-4.

[0018] However, these methods are not suitable for the intended industrial purposes because they involve using pre-tensioned yarn supplied according to regulations.

[0019] Finally, some solutions are given, for example, in patent documents CN-A-111235709, CN-A-109389583, CN-A-105386174, DE102018111648A1, WO2019130209A3, JP2018178282A and DE102016121662A1.

[0020] Furthermore, the solutions known to date are unsatisfactory in terms of the reliability of the results. In fact, slight irregularities in the fibers, such as a slight increase in fiber size in a region, are often misidentified as tangles or other defects. Summary of the Invention

[0021] The purpose of this invention is to provide a method and apparatus for detecting defects in yarn processed in a spinning machine, which meets industrial requirements and overcomes the disadvantages discussed above with reference to the prior art.

[0022] This objective is achieved by the method and apparatus according to embodiments of the present invention. Attached Figure Description

[0023] The features and advantages of the method and apparatus according to the invention will become apparent from the following description, given by way of non-limiting example with reference to the accompanying drawings, wherein:

[0024] - Figure 1 A spinning machine equipped with a detection device according to the present invention is shown;

[0025] - Figure 2 yes Figure 1 A schematic diagram of a spinning machine;

[0026] - Figure 3a and Figure 3b A positive image of the original cotton knot with the corresponding pixel outline is shown;

[0027] - Figure 4 A frontal image of a set of synthetic cotton knots is shown. Detailed Implementation

[0028] For clarity, reference will be made below to methods and equipment for detecting cotton knots; however, it should be understood that the present invention is generally applicable to defect detection.

[0029] Referring to the accompanying drawings, 1 generally represents a spinning machine for a spinning production line that obtains yarn spools from roving bobbins, the spinning machine having an extension along the longitudinal axis X.

[0030] The spinning machine 1 includes a frame 2 for supporting components and a warp beam 4 supported by the frame 2, the frame 2 being made of one or more components arranged side by side.

[0031] The warp beam frame 4 includes vertical columns 6 and multiple longitudinal transverse members 8, which are supported by the columns 6 and positioned at a predetermined height. The transverse members 8 are used to support multiple suspended bobbins B of the roving.

[0032] Below the crossbar 8, that is, downstream of the bobbin B, the spinning machine 1 includes a drafting device 10 supported by the frame 2.

[0033] The drawing device 10 includes a plurality of lower drawing cylindrical members 12a-12d, typically three or four. The lower drawing cylindrical members are motorized and extend longitudinally, are made of one or more parts, have different structures, and are arranged side by side and aligned.

[0034] The drafting device 10 also includes a plurality of pressure arms 14 arranged longitudinally side by side. Each pressure arm carries an idle pressure roller 16.

[0035] The pressure roller 16 is connected to the pressure cylindrical members 12a-12d to form a drafting pair. The roving passes through the drafting pair and is drafted by the circumferential speed of each drafting pair, which increases from upstream to downstream.

[0036] The spinning machine 1 also includes a winding device 17 arranged below the drafting device 10 and immediately downstream of the first drafting cylindrical member 12a.

[0037] The winding device 17 includes a yarn guiding assembly 18, which includes a support 20 connected to the frame 2 and a plurality of yarn guides 22 supported by the support 20 and arranged longitudinally side by side.

[0038] The winding device 17 also includes a guide rail 23, which is located below the yarn guiding assembly 18, i.e. downstream of the yarn guide 22. The guide rail 23 is supported by the frame 2 and is capable of vertical movement in a reciprocating motion.

[0039] The winding device 17 also includes a plurality of spindles 24 arranged side by side along the longitudinal direction of the track 23, each spindle being able to rotate about its own vertical axis.

[0040] During normal operation of the spinning machine 1, the roving wound in a predetermined bobbin B travels through a first path segment to enter the drafting device 10, and the roving leaving the drafting device is drafted; the drafted roving R travels through a second path segment Sr between the drafting device and the corresponding spindle 24, passing through the corresponding yarn guide 22. The yarn obtained from the drafting and twisting of the roving is wound onto a tube fitted onto the spindle to form a yarn bobbin.

[0041] According to the present invention, a detection area is defined between the drafting device 10 and the winding device 17, through which a segment Sr of the second path of the drafted roving R passes.

[0042] The stretched roving passing through the inspection area undergoes a process for defect detection, particularly for detecting tangles or knots commonly referred to as "neps".

[0043] The detection process includes a collection step, during which at least one segment R of the roving R passing through the detection area during transmission is sampled. Images were acquired.

[0044] For this purpose, the spinning machine 1 includes a data acquisition device 60, such as a digital data acquisition device 60, which includes a camera, and the data acquisition device 60 is adapted to acquire at least one segment R of the roving R passing through the detection area during transmission. Image I was acquired.

[0045] Furthermore, preferably, the spinning machine 1 includes an irradiation device 62, which is adapted to irradiate the section R of the detection area, including the roving R, for example by warm light or cold light, or in a variant embodiment by infrared light. At least one area is irradiated.

[0046] The detection process also includes a step performed by the processing device 70 to process the image I acquired by the acquisition device 60.

[0047] The processing device 70 is configured to detect knots using an adaptive or machine learning detection algorithm, particularly a Haar cascade type, preferably based on the Viola-Jones method. The Viola-Jones method is described in the paper "Rapid object detection using a boosted cascade of simple features" published by Paul Viola and Michael Jones at the Conference on Computer Vision and Pattern Recognition (2001), and their teachings on the algorithm implementation are specifically incorporated herein.

[0048] The detection algorithm is based on groups of positive images Ip and groups of negative images In, where the positive image Ip is the segment R of the drafted roving R. An image with knots, the negative image In showing the segment R of the stretched roving R. It does not have knots.

[0049] According to the first embodiment (referred to as "having original neps"), learning begins with an image of a drawn roving, for example, an image acquired during normal use of a predetermined spinning machine, thus depicting segments with neps and segments without neps. A positive image Ip in a group of positive images is determined by processing the pixel profiles of each image and selecting images whose pixel profiles have at least one peak exceeding a threshold as positive images. The pixel profiles are obtained by adding bright pixels to each row of the image, and the threshold is determined, for example, by the average value and standard deviation. Figure 3a and Figure 3b ).

[0050] Furthermore, preferably, the positive image is further selected to eliminate peaks caused by individual worn hairs; this further selection is performed either by an additional selection algorithm or manually.

[0051] According to another embodiment (referred to as "with synthetic knots"), the positive image Ip is digitally constructed and consists of a semi-circular or semi-elliptical image, preferably vertical. Figure 4 For example, the lengths of the minor axis and the major axis are different from each other.

[0052] According to another embodiment (referred to as "hybrid"), the positive image Ip includes a positive image with original knots and a positive image with synthetic knots, that is, a combination of the two embodiments described above.

[0053] According to another embodiment of the invention, the processing device is configured to detect cotton knots using adaptive or machine learning detection algorithms of the type of convolutional neural network (CNN) and / or recurrent neural network (RNN).

[0054] During normal operation of the spinning machine, when the roving R passes through the detection area during transmission, the acquisition device 60 continuously acquires segments R of the roving R. Image I.

[0055] The image I is processed by the processing device 70 using a machine learning detection algorithm or using a convolutional neural network (CNN) and / or a recurrent neural network (RNN) to detect the frequency of knots and preferably the shape of knots, thus classifying them by type. The machine learning detection algorithm is particularly of the Haar cascade type, preferably based on the Viola-Jones method.

[0056] Based on these findings, as part of methods for managing spinning machines or spinning production lines that include spinning machines and their upstream machines such as carding machines, combing machines, drawing frames, and roving frames, the processing parameters of spinning machines, such as yarn twisting, roving drafting and / or pre-drafting, roving type and weight, shaft rubber hardness, bobbin specifications, shaft pressure, and production speed, are adjusted to improve roving quality, or the parameters of upstream processes of the spinning machine are adjusted, such as twisting, drafting and / or pre-drafting, shaft rubber hardness, bobbin specifications, shaft pressure, roving frame production speed, production speed and drafting assembly specifications in the drafting machine, scrap percentage, number of strokes in the combing machine and drafting unit specifications in the combing machine, production speed, nipple and impurity removal in the carding machine and opening and cleaning line, or maintenance work performed on the spinning machine or its upstream machines.

[0057] Innovatively, the apparatus and method for detecting defects in spinning machines according to the present invention meet the needs of the industry and overcome the aforementioned defects.

[0058] In fact, since the testing is carried out continuously on the roving processed in the spinning machine, interventions can be made to modify the processing parameters on the spinning machine or upstream machines, or through maintenance interventions, in order to improve yarn quality.

[0059] Furthermore, advantageously, the present invention ensures good reliability of the results because it allows for the differentiation between industrially acceptable major defects and other minor irregularities.

[0060] Furthermore, the tests conducted showed a good correspondence between the readings obtained using the device of the present invention and the tests based on commonly used testing instruments as described above.

[0061] Furthermore, it is advantageous that the image processing according to the invention is very fast and enables continuous defect detection and rapid action to improve production.

[0062] Obviously, those skilled in the art can modify the above methods and apparatus to meet possible needs, and all such modifications are included within the scope of protection defined by the appended claims.

Claims

1. A method for detecting defects in drafted roving (R) processed in a spinning machine, the method comprising the steps of: An image (I) of the drafted roving (R) being transported in a segment (Sr) of the second path of the roving (R) is acquired, and the image (I) is digitally processed to detect the defect, wherein the image (I) processing step involves using a machine learning detection algorithm to detect the defect, wherein the machine learning detection algorithm is learned based on a set of positive images (Ip) and a set of negative images (In), wherein the roving segment depicted in the positive image has a defect and the roving segment depicted in the negative image does not have the defect.

2. The method according to claim 1, wherein, The processing steps provide a step of detecting the frequency of defects along the roving (R).

3. The method according to claim 1 or 2, wherein, The processing steps provide steps for detecting the type of the defect.

4. The method according to claim 1 or 2, wherein, Starting with an image of the stretched roving, defective and non-defective segments are learned. The positive image (Ip) is determined by processing the pixel profile of each image and selecting images whose pixel profiles have at least one peak exceeding a threshold as positive images, wherein the pixel profiles are obtained by adding bright pixels to each row of the image.

5. The method according to claim 4, wherein, The threshold is determined by the average value and standard deviation of the pixel contour.

6. The method according to claim 4, wherein, The learning is performed during the normal use of the predetermined spinning machine to acquire images of the roving.

7. The method according to claim 4, wherein, The positive image is subjected to further selection by eliminating the peaks formed by individually pulled-out hairs.

8. The method according to claim 7, wherein, The additional selection is performed using an alternative selection algorithm.

9. The method according to claim 7, wherein, The other option mentioned is to be performed manually.

10. The method according to claim 1 or 2, wherein, The positive image (Ip) is digitally constructed and formed from images.

11. The method according to claim 1 or 2, wherein, The positive image (Ip) is digitally constructed and formed by a vertical semicircle or semiellipse image.

12. The method according to claim 10, wherein, The image differs from each other in the length of its minor axis and the length of its major axis.

13. The method according to claim 11, wherein, The vertical semicircle or semiellipse image differs from each other in the length of its minor axis and major axis.

14. The method according to claim 1 or 2, wherein, A portion of the positive image (Ip) was obtained through a machine learning detection algorithm; and The remaining portion of the positive image (Ip) is obtained through digital construction.

15. The method according to claim 1 or 2, wherein, The machine learning detection algorithm is of the Haar cascade type.

16. The method according to claim 12, wherein, Haar cascade-type machine learning detection algorithms are based on the Viola-Jones method.

17. The method according to claim 1 or 2, wherein, The machine learning detection algorithm has the Convolutional Neural Network (CNN) type and / or Recurrent Neural Network (RNN) type.

18. A method for managing a spinning machine (1) in a spinning production line, comprising: The method for detecting defects in drafted roving (R) processed in a spinning machine according to any one of claims 1 to 17, and A series of steps to change the processing parameters of the spinning machine or to perform maintenance operations on the spinning machine.

19. A method for managing a spinning production line, the spinning production line comprising a spinning machine (1) and a fabric handling machine upstream of the spinning machine, the method comprising: The method for detecting defects in drafted roving (R) processed in a spinning machine according to any one of claims 1 to 17, and A series of steps including changing the processing parameters of at least one of the machines upstream of the spinning machine or performing maintenance operations on at least one of the machines upstream of the spinning machine.

20. An apparatus for detecting defects in drafted roving (R) processed in a spinning machine, the apparatus comprising: A data acquisition device (60) is adapted to acquire an image (I) of the drafted roving (R) from a detection area located between the drafting device (10) and the winding device (17) of the spinning machine, the detection area being traversed by a segment (Sr) of a second path of the roving (R). A processing device (70) is operably connected to the acquisition device for digitally processing the image (I) and detecting defects; The processing device (70) is configured to detect defects by means of a machine learning detection algorithm, wherein the machine learning detection algorithm is learned based on a set of positive images (Ip) and a set of negative images (In), wherein the roving segment represented in the positive image has a defect and the drafted roving segment represented in the negative image does not have the defect.

Images