Method and apparatus for classifying events in elongate testing material

Abstract

In this computer-implemented method for classifying an event in an elongate textile test material, at least one measured value of at least one event parameter is measured (901) for the event. At least one reference data set is provided (902) containing, for a respective reference event, at least one reference value of the at least one event parameter and a classification of the reference event. The at least one measured value is compared with the at least one reference value of all the reference data sets (904). The event is classified (908) according to the classification of the reference event, at least one reference value of the reference event being most similar to the at least one measured value according to a predefined similarity criterion. The method achieves high quality and at the same time high productivity.

Description

Technical Field

[0001] This invention relates to the field of textile quality control. Specifically, it relates to a computer-implemented method and apparatus for classifying events in long, thin textile test materials. The invention also relates to a yarn cleaning system and a yarn processing machine. The computer-implemented method and apparatus are preferably (but not limited to) used in yarn cleaners on spinning machines or winding machines.

[0002] Furthermore, the present invention relates to a computer-implemented method for providing at least one reference dataset to a method or apparatus for classifying events according to the present invention. Finally, the present invention also relates to a computer-readable medium having stored thereon at least one reference dataset provided according to the method. Background Technology

[0003] To ensure yarn quality, a so-called yarn cleaner is used on spinning or winding machines. A yarn cleaner includes a measuring head with at least one sensor that scans the moving yarn. Commonly used sensor principles are capacitive and optical, both described in WO-2012 / 051730 A1. The purpose of scanning is to detect events in the yarn.

[0004] In this document, an "event" in a long, thin textile test material is defined as a confined location in the longitudinal direction (i.e., a finite length typically less than 1 m) where at least one specific measurement deviates from the corresponding target value. The event passes through a fixed point, such as a sensor, within a finite time interval (typically less than 1 s) as the long, thin textile test material is moved longitudinally. Examples of events include thick places, thin places, or foreign objects in the yarn.

[0005] Measurements obtained from sensor signals are continuously evaluated using specified evaluation criteria (such as cleanup limits). If a yarn event is below the cleanup limit, it is tolerable; if it is above the cleanup limit, it is an intolerable yarn defect that must be removed from the yarn or at least registered. Therefore, yarn cleanup is based on classifying yarn events into a two-category system: tolerable yarn events and intolerable yarn events.

[0006] WO-2011 / 038524 A1 discloses a method for determining cleanup limits. First, a statistical representation of the yarn is determined by measuring the yarn. Based on this statistical representation, cleanup limits are calculated and proposed for the application, wherein the expected number of unacceptable events related to this cleanup limit is calculated and output. The operator can comment on the expected number of unacceptable events, and then the cleanup limits are automatically set based on the comments.

[0007] Yarn events can be input as points into a two-dimensional classification scheme, in which error length is typically plotted along the horizontal axis, and error magnitude (deviation from the target value for factors such as mass per yarn length, yarn cross-section, and yarn reflectivity) is plotted along the vertical axis. Examples of cleanup limits and classification schemes are also given in WO-2011 / 038524 A1.

[0008] According to EP-0'685'580 A1, yarn defects are recorded, entered into a classification scheme, and counted in each category. This generates an error pattern in the classification scheme. Comparing this error pattern with a specified model reference pattern allows conclusions to be drawn regarding the causes of the recorded errors. The specified reference pattern can be determined through preliminary research or derived from experience.

[0009] According to WO-2010 / 078665 A1, the density of events in the classification scheme is determined by the size and length of defects. In the classification scheme, the yarn body is represented as a region. A region is defined on one side by an abscissa and on the other side by a ordinate, and further defined by lines in the event field that follow a constant event density. Events occurring within the yarn body can be considered statistical "noise" belonging to the actual yarn and should not be removed from the yarn.

[0010] WO-2013 / 185246 A1 also teaches that a first density line associated with the first test material should represent a constant event density in the classification scheme. In addition to the first density line, a reference density line is also shown, which follows the same event density as the first density line but is at least partially associated with a reference test material different from the first test material. This allows for a comparison of the curves of the first density line and the reference density line. Comparison of the progression of the density lines allows for a comparison of the quality of the first test material and the reference test material.

[0011] WO-00 / 73189 A1 sets the following objective for itself: to allow for improvements, simplification, and rapid adjustments to cleaning limits so that the impact of cleaning limits on the final product can be predicted more accurately. To this end, a representation of defects in the final product (e.g., in yarn) will be generated based on the cleaning limits, making the impact of defects in the final product visible.

[0012] Spinning practice shows that even with seemingly optimally set cleaning limits, yarn cleaning can sometimes be unsatisfactory. On the one hand, after yarn cleaning, there may still be events in the yarn that are below the cleaning limit but still considered destructive. To remove them, the cleaning limit can be "tightened," i.e., shifted downwards. However, this will lead to tolerable yarn events being unnecessarily rejected, thus reducing productivity. On the other hand, yarn cleaning may remove events from the yarn that, upon closer inspection, are found to be completely non-destructive. To prevent this, the cleaning limit can be "relaxed," i.e., moved upwards. However, this will mean that other destructive yarn defects will not be rejected, thus reducing quality.

[0013] WO-97 / 31262 A1 relates to the electronic simulation of fabric patterns. It proposes imaging yarn segments with a video camera, storing the images digitally, and weaving them together to obtain a realistic image of the resulting fabric. The simulated image can be output via screen or printer. Summary of the Invention

[0014] One object of the present invention is to provide a method and apparatus for classifying events in an elongated textile test material moving along its longitudinal direction, which reduces or eliminates the disadvantages of the prior art. Specifically, as many destructive yarn defects as possible should be classified as destructive, while permissible yarn events should not be unnecessarily classified as destructive, and as many permissible yarn events as possible should be classified as permissible, while destructive yarn defects should not be classified as permissible. In this way, both high quality and high productivity can be achieved simultaneously.

[0015] Another object of the present invention is to provide a method and system for cleaning yarn that reduces or eliminates the disadvantages of the prior art. In particular, it will increase the productivity of the yarn cleaning process and the quality of the cleaned yarn.

[0016] Furthermore, a method for providing at least one reference dataset for the method or apparatus according to the present invention will be provided.

[0017] These and other objectives are achieved by the method and apparatus according to the invention, as defined in the independent patent claims. The dependent patent claims describe a method and system for yarn cleaning, as well as advantageous embodiments.

[0018] This invention is based on the consideration that the quantitative assessment of yarn events does not always correspond to the assessment made by the human eye. Algorithms used to date for yarn cleaning (which compare the results of physical measurements of the yarn with predetermined cleaning limits) yield satisfactory results in many cases, but cannot handle the aforementioned exceptions. In many situations, only humans can conclusively determine whether a yarn event is considered disruptive. After all, the final textile product made from said yarn is also observed by humans.

[0019] Based on this consideration, the present invention achieves its purpose by having yarn events assessed and classified by the operator. Furthermore, yarn events are measured using sensors known from the prior art. Classifications are assigned to the measurements associated with the yarn events and stored together with the measurements as a reference dataset. Preferably, this process is repeated for several different yarn events, thereby constructing and providing a database of reference datasets.

[0020] The computer-implemented method according to the present invention is used for classifying events in slender textile test materials. The method includes the following steps: a) Measure at least one value of at least one event parameter for the event; b) Provide at least one reference dataset, wherein each reference dataset contains at least one reference value for at least one event parameter and a classification of the reference event for the reference event; c) Compare at least one measurement with at least one reference value from all reference datasets; and d) Classify the events according to the classification of reference events, wherein at least one reference value of the reference event is most similar to at least one measurement value according to a predetermined similarity criterion.

[0021] In one implementation, at least one event parameter is the number of events, such as the deviation of the length-dependent mass density, cross-sectional area, or reflectance of the test material from a target value, and at least one reference value and at least one measurement value are each signals consisting of multiple values ​​of the number of events varying with length, location, or time. In this case, the similarity criterion may include the cross-correlation of the two signals.

[0022] In another embodiment, at least one event parameter includes the number of events (e.g., the deviation of the test material's length-related mass density, cross-sectional area, or reflectance from the target value) and the event length. In this case, the similarity criterion may include the distance between a point representing an event and a point representing a reference event in a coordinate system defined by the event length and event amplitude.

[0023] In one implementation, several different event parameters are considered. At least one first event parameter can be measured capacitively, for example, and at least one second event parameter can be measured optically, for example.

[0024] In one embodiment, measurements of at least one event parameter are taken while the slender textile test material is moved along its longitudinal direction.

[0025] At least one reference dataset can be provided by: measuring at least one reference value for each reference event, visually evaluating and classifying the reference events by an operator, and assigning the classification to at least one reference value and storing it together with the reference value as a reference dataset. The reference events are preferably visually evaluated by: capturing images of the reference events and presenting the images to the operator for visual evaluation.

[0026] In one implementation, the classification is performed in a classification system having exactly two categories: a first category with permissible events and a second category with non-permissible events.

[0027] In the computer-implemented method according to the invention for cleaning yarn moving along its longitudinal direction, events in the yarn are classified into the two categories mentioned above, and events classified into the second category are removed from the yarn.

[0028] The apparatus according to the invention for classifying events in slender textile test materials includes a) A measuring device for measuring at least one measured value of at least one event parameter for an event; and Computer systems, which have b) A memory for storing at least one reference dataset, wherein each reference dataset contains at least one reference value for at least one event parameter and a classification of the reference event for the reference event; c) A processor configured to compare at least one measurement with at least one reference value from all reference datasets; d) A processor configured to classify events according to a classification of reference events, wherein at least one reference value of the reference event is most similar to at least one measurement value according to a predetermined similarity criterion.

[0029] In one embodiment of the apparatus, at least one event parameter is the number of events, such as the deviation of the length-related mass density, cross-sectional area, or reflectance of the test material from a target value; at least one reference value and at least one measured value are each signals consisting of a plurality of values ​​of the number of events varying with length, location, or time; and the measuring device is configured to receive said signals. At least one event parameter may include the number of events (e.g., the deviation of the length-related mass density, cross-sectional area, or reflectance of the test material from a target value) and the event length.

[0030] In one embodiment, the measuring device is designed to measure several different event parameters. The measuring device may, for example, be designed to perform capacitive measurements of at least one first event parameter and optical measurements of at least one second event parameter.

[0031] In one implementation, the processor is configured to perform classification in a classification system having exactly two categories: a first category with permissible events and a second category with non-permissible events.

[0032] The yarn cleaning system according to the invention for cleaning yarns moving along their longitudinal direction comprises: a device having a processor for classifying in a classification system having the two categories described above; and a cutting device that can be triggered by a computer system to cut the yarns upon classifying the event as a second category.

[0033] The yarn processing machine (e.g., a winding machine or a spinning machine) according to the invention comprises: a plurality of yarn processing stations, wherein at each yarn processing station, yarn is wound onto a bobbin; and the yarn cleaning system according to the invention described above.

[0034] In one embodiment, the yarn processing machine according to the invention comprises: at least one camera disposed at one of the yarn processing stations to capture images of events in the yarn; an output unit connected to the at least one camera to output the images captured by the camera; and an input unit connected to a processor to input and output image classification.

[0035] According to the invention, a computer-implemented method for providing at least one reference dataset for the method or apparatus according to the invention comprises, for each reference event in a slender textile test material, measuring at least one reference value for at least one event parameter for the reference event, visually evaluating and classifying the reference event by an operator, and assigning the classification to at least one reference value and storing it together with the at least one reference value as a reference dataset.

[0036] In one implementation, a reference event is visually assessed by capturing an image of the reference event and presenting the image to the operator for visual assessment.

[0037] The present invention also relates to a computer-readable medium having stored thereon at least one reference dataset provided according to the present invention.

[0038] One advantage of this invention is that events in the test material are categorized based on human perception rather than abstract parameters. This helps ensure that the final textile product made from the test material is free of any defects that would be considered disruptive to the human eye. In terms of yarn cleaning, removing all defects considered disruptive from the yarn guarantees high yarn quality. Simultaneously, avoiding the removal of non-disruptive yarn events increases productivity. Attached Figure Description

[0039] The invention will now be described in detail with reference to the accompanying drawings. Some exemplary embodiments relate to a yarn cleaner at the winding station of a winding machine. However, this should not limit the generality of the invention. The yarn cleaner is used not only on winding machines but also on spinning machines. However, the invention is not limited to applications of yarn cleaning. The invention is also applicable to other slender textile test materials, such as slivers.

[0040] Figure 1 A schematic diagram of a yarn winding machine with a yarn cleaning system is shown.

[0041] Figure 2 shows (a) the virtual measurement signal and (b) the virtual reference signal as the length and position change.

[0042] Figure 3 shows (a) the virtual mass signal and (b) the virtual diameter signal as the length position varies.

[0043] Figure 4 A diagram illustrating the similarity criteria for the two signals from Figure 3 is shown.

[0044] Figures 5 through 7 each show (a) four different measurement signals for one and the same yarn event and (b) a photograph of the yarn event.

[0045] Figure 8 A schematic representation of an event field is shown, comprising a yarn body, a cleaning curve, and various reference events.

[0046] Figure 9 A flowchart of one embodiment of a method for classifying events according to the present invention is shown.

[0047] Figure 10 A flowchart of one embodiment of a method for providing at least one reference dataset according to the present invention is shown.

[0048] Figure 11 A schematic representation of a table in a database with a reference dataset is shown. Detailed Implementation

[0049] Figure 1 A true schematic representation of a yarn winding machine 100 according to the invention, having a plurality of winding stations 101, is shown. At each winding station 101, yarn 110 is wound from bobbin 111 onto a cross-wound bobbin 112 during the winding process. In this example, the yarn 110 moves from bottom to top along its longitudinal direction, as indicated by arrow 113. A device 120 according to the invention is installed in the winding machine 100.

[0050] At each winding position 101, the yarn 110 is monitored by a measuring head 121 of the device 120 according to the invention. The measuring head 121 measures at least one measured value of at least one event parameter for events in the yarn 110. For this purpose, the device may be equipped with capacitive, optical and / or other sensors, as well as an evaluation unit for evaluating the sensor signals. Such measuring heads 121 are known in themselves and need not be further explained here.

[0051] Preferably, a cutting device is also provided at each winding station 101. This device is used to remove yarn segments with unacceptable defects from the yarn 110. The cutting device receives the corresponding cutting command from the evaluation unit of the measuring head. The cutting device is preferably integrated into the measuring head 121.

[0052] All measuring heads 121 are connected via data lines 122 to the central control unit 123 of the device 120 according to the invention. The measuring heads 121 are set and controlled by the control unit 123, and the measuring heads transmit data such as measured values ​​to the control unit 123. The data lines 122 may be designed as, for example, a serial bus. The control unit 123 is equipped with a central computer 124, which is used, on the one hand, to collect and further process the measurement data from the measuring heads 121, and on the other hand, to adjust the measuring heads 121. The central computer 124 is connected to or contains a database 125. Furthermore, the control unit 123 has an output unit 126 and an input unit 127 for the operator. The output unit 126 may be designed as, for example, a screen and / or a printer. The input unit 127 may include, for example, a keyboard or a computer mouse. The output unit 126 and the input unit 127 may be designed together as a sensor screen. The output unit 126 and the input unit 127 may be designed as a mobile terminal device wirelessly connected to the central computer 124.

[0053] Figure 2(a) illustrates a virtual measurement signal 210 for a yarn event, as it can be measured by the measuring head 121. The measurement signal 210 consists of numerous measurements recorded along the length position L (horizontal axis 201) of the yarn 110, representing the event parameter ΔM (vertical axis 202). The event parameter ΔM can be, for example, the deviation of the length-related mass density M of the yarn 110 from a capacitive measurement of a target value. Another example of an event parameter is the deviation of the yarn diameter from an optical measurement of a target value. The target value is preferably determined by calculating a sliding average relative to multiple measurements. The measurements can be recorded, for example, every 2 mm and combined to form the measurement signal 210 according to Figure 2(a).

[0054] In a representation similar to Figure 2(a), Figure 2(b) shows a virtual reference signal 220 for a reference event. The reference signal 220 is then composed of a number of reference values ​​recorded along the length L of the same event parameter ΔM. The reference yarn may be different from or the same as the yarn whose measurement signal 210 is recorded in Figure 2(a). If the reference yarn is different, it should have similar properties to achieve meaningful classification of yarn events; in particular, it should be made of the same material, have the same number of yarns, and be manufactured using the same spinning process.

[0055] It should be understood that signals 210 and 220 can be represented in terms of time rather than length position L. If the instantaneous speed of the yarn is known, it is easy to convert each of the two signals 210 and 220 from length dependence to time dependence, and vice versa.

[0056] In the method according to the invention (which is in) Figure 9 In the full discussion below, the measured signal 210 is compared with the reference signal 220. The two signals 210 and 220 can be compared, for example, using a cross-correlation function known to those skilled in the art. The maximum value of the cross-correlation function is a measure of the similarity between the two signals 210 and 220: the larger the maximum value, the more similar the signals. Furthermore, those skilled in the art are familiar with other methods for comparing the two signals 210 and 220.

[0057] Another method for comparing the measured signal 210 with the reference signal 220 utilizes the relatively simple basic form of yarn events, which are typically either "peaks" or "valleys." This method simplifies the two signals 210 and 220 to a single value for each of two event parameters. The first event parameter is the average event size ΔM. M or ΔM R For example, the average of all signal values ​​that reach at least 80% of the signal's maximum value. The second event parameter is the event length L. M or LR For example, the length of an event at 50% of its average event height. This means that each event can be represented by a two-dimensional vector (L... M , ΔM M ) or (L R , ΔM R The difference vector is used to represent the similarity between two events. The length of the difference vector is a measure of the similarity between the two events: the shorter the length, the more similar the two events. This comparison method can be generalized to more than two dimensions (event parameters).

[0058] Figure 3 shows two different virtual measurement signals 310 and 320 for a single yarn event, varying with its length position L (horizontal axis 301). The first measurement signal 310, according to Figure 3(a), can be composed of the measured value of the deviation ΔM (vertical axis 302) of the yarn's length-related mass density M from the target value, as shown in Figure 2(a). The second measurement signal, according to Figure 3(b), can be composed of the measured value of the deviation ΔD (vertical axis 303) of the yarn's cross-section D from the target value.

[0059] Each of the two measurement signals 310, 320 can be compared individually with a corresponding (not shown) reference signal, for example, using a cross-correlation function as discussed above. Figure 4 The figure shows the two maximum values ​​R of the cross-correlation function. 最大 (ΔM) (horizontal axis 401) and R 最大 (ΔD) (vertical axis 402) can be input as the coordinates of point 410 in the two-dimensional coordinate system. The distance 411 between point 410 and the ideal corner point P(1,1) is a measure of the similarity between the two measurement signals 310 and 320 in Figure 3 and the corresponding reference signal: the shorter the distance, the more similar they are.

[0060] Alternatively, each of the two measurement signals 310 and 320 from Figure 3 can be simplified to a few measurements of fewer event parameters, as discussed above in conjunction with Figure 2. Each yarn event can then be represented, for example, by a four-dimensional vector, the components of which are the event size and event length determined from the two measurement signals 310 and 320. The length of the difference vector between the measurement vector and the reference vector is a measure of the similarity between the two events: the shorter the length, the more similar they are.

[0061] As a supplement to the virtual measurement signals 210, 310, and 320 in Figures 2 and 3, Figures 5 to 7 refer to real yarn events, which can be events to be classified or reference events. In the sub- Figure 5(a) , 6(a)In 7(a), actual measurement signals showing the variation of the number of four events (vertical axes 502, 602, 702) with their length position or time (horizontal axes 501, 601, 701) are shown. The four measurement signals are as follows: • Solid lines: Measurement signals 511, 611, and 711 of the capacitive sensor for mass deviation; • Fine dashed lines: Inverted measurement signals 512, 612, and 712 of an optical sensor that measures light transmitted through the yarn for diameter deviation; • Dashed line: Measurement signals 513, 613, 713 of an optical sensor that measures green light reflected from yarn for foreign objects; • Rough dashed lines: Measurement signals 514, 614, and 714 for foreign objects by optical sensors that measure red light reflected from yarn.

[0062] In the method according to the invention, any one of the four measurement signals, any combination of two or three measurement signals, or all four measurement signals can be considered.

[0063] son Figure 5(b) , 6(b) Figures 520, 620, and 720 (i.e., photographs) of corresponding yarn events 540, 640, and 740 in yarns 530, 630, and 730, respectively, but not necessarily in the order of the sub-yarn events 540, 640, and 740. Figure 5(a) , 6(a) The proportions are the same as those in 7(a), and may be offset relative to the measurement signal along the longitudinal axes 501, 601, and 701. The vertical lines in the background simply indicate the proportions (10 mm) on the yarn sheet used.

[0064] Table 1 provides an overview of the three yarn events 540, 640, and 740 and some of their measurements.

[0065] Table 1 Regarding the cotton knot 540 in Figure 5 and the short, thick nodule 640 in Figure 6, the quality signals 511 and 611 and the diameter signals 512 and 612 show significant deviations, while the two foreign object signals 513, 514, 613, and 614 show relatively insignificant deflections. The opposite is true for the foreign fiber 740 in Figure 7.

[0066] The example in Figure 7 illustrates an advantage of the invention. It is assumed that the method according to the invention is performed using only the diameter signal 712, and the other three signals 711, 713, and 714 are unavailable. The diameter signal 712 is triangular, with each peak corresponding to the twisting of the foreign fiber 740. In conventional yarn cleaning, these may be below the cleaning threshold, meaning the foreign fiber 740 will not be removed. However, using the method according to the invention, image 720, and therefore the reference event in Figure 7, is classified as unacceptable. Measurement signals with similar triangular curves are also classified as unacceptable according to the invention and removed from the yarn during yarn cleaning.

[0067] Yarn events can be represented using known event fields 800, such as... Figure 8 As illustrated schematically, event field 800 is a two-dimensional Cartesian coordinate system. The values ​​of the number of events (e.g., the deviation ΔM of the yarn length-related mass density M from the target value) are plotted along the vertical axis 802, and the values ​​of the event length L are plotted along the horizontal axis 801 (see Figure 2). Yarn events, i.e., measured values ​​and / or reference values, can be entered as points into this event field 800. Figure 8 The points 811 and 812 entered in the example are reference values.

[0068] exist Figure 8 The event field 800 inputs cleanup limits in the form of a cleanup curve 830. The cleanup curve 830 divides the event field 800 into two regions 831 and 832: a first region 831 where deviations are tolerated; and a second region 832, which is complementary to the first region 831, where deviations are eliminated from the yarn, or at least registered as unacceptable errors. Therefore, the cleanup curve 830 specifies how far a yarn event of a given length can deviate from the target value, or how long a yarn event with a given deviation can remain tolerated.

[0069] Furthermore, as shown in WO-2010 / 078665 A1, yarn body 840 is... Figure 8 In event field 800, to determine yarn body 840, sufficiently long yarn segments are measured during calibration. The density of events in event field 800 is determined from the measurements of the corresponding associated event parameters Δ, M, and L. This allows for a unique assignment of event density to each point in event field 800. A sufficiently high limit event density is selected, for example, 1000 events per 100 km of yarn length. Connecting all points in event field 800 to which the limit event density is assigned yields density curve 841, which separates yarn body 840 from the rest of event field 800. Yarn body 840 is defined by coordinate axes 801 and 802 themselves oriented towards the two coordinate axes 801 and 802. These boundaries form contiguous region 840, which is characteristic of the yarn.

[0070] The calculation and representation of the yarn body 840 according to the present invention enables the operator to quickly and intuitively grasp the characteristics of the yarn being inspected, and thereby determine the cleaning curve 830. The event density in the yarn body 840 is so high that attempting to clean events from the yarn body 840 is meaningless. Therefore, the cleaning curve 830 should be located above the yarn body 840.

[0071] exist Figure 8 In event field 800, two types of reference events 811 and 812 are marked as points. The first type 811 is located in the area between yarn body 840 and cleanup curve 830. Yarn events should generally not be rejected in this area below cleanup curve 830. However, it is still possible that some events in this area are considered destructive and therefore need to be rejected. Moving cleanup curve 830 downwards would result in the unnecessary rejection of tolerable yarn events, thus reducing productivity. The present invention takes a different approach by defining certain reference events 811 within said area. Yarn events similar to the first type of reference event 811 are rejected as exceptions to the general rule. Figure 8 In the example, points corresponding to the layers of the first type 811 reference event are marked with a cross. To establish a connection with Figure 2(b), coordinates L are plotted for these points. R , ΔM R .

[0072] The second type of reference event 812 is located above the cleaning curve 830. This is where normally destructive yarn defects that should be rejected should be located. However, there may be some yarn events 812 above the cleaning curve 830 that, upon closer inspection, are found to be non-destructive at all. Shifting the cleaning curve 830 upwards would result in other destructive yarn defects not being rejected, thus reducing quality. In this case, the invention also provides for an exception to the general rule, namely, certain reference events 812 above the cleaning curve 830 that are not destructive at all. Figure 8 The circle marks the yarn events. Yarn events similar to these reference events 812 of the second type should be excluded as exceptions to the general rules.

[0073] Figure 9 The flowchart in the diagram now illustrates a computer-implemented method according to the invention for classifying events in a slender textile test material moving along its longitudinal direction.

[0074] For the event to be classified, at least one measurement value of at least one event parameter 901 is measured. A first example of an event parameter is the mass deviation ΔM discussed in conjunction with Figure 2, from which multiple measurements of the mass deviation are combined to form a measurement signal 210. A second example is two event parameters also discussed in conjunction with Figure 2: event length L and mass deviation ΔM. Further examples are capacitive and optical measurement results discussed in conjunction with Figure 3, where length-dependent signals 310, 320 and / or deviation ΔM or ΔD, along with length L, can be used as event parameters.

[0075] Provide at least one reference dataset (902) and preferably several reference datasets, preferably in a reference database. Each reference dataset relates to a reference event and contains at least one reference value for at least one event parameter and a classification of the reference event. In this context, "classification" refers to assigning exactly one category to the reference event. (See below for reference...) Figure 10 The provision of the reference dataset will be discussed in more detail in section 902.

[0076] Retrieve reference datasets one by one from the reference database. Compare at least one measurement with at least one reference value in the corresponding reference dataset. If at least one measurement is most similar to at least one reference value in all previous reference datasets, then, for example, in temporary storage, classification is made available from the relevant reference datasets. An example of a similarity criterion is given above. Repeat the comparison of at least one measurement with at least one reference value until all reference datasets have been verified.

[0077] Following comparison 904, the currently provided classification is a classification of a reference event, wherein at least one reference value of the reference event is most similar to at least one measurement value according to a specified similarity criterion. The event is then classified according to the available classifications 908.

[0078] If the yarn cleaning method according to the invention is used, the classification 908 is preferably performed in a classification system having exactly two categories: a first category with permissible events and a second category with non-permissible events. Events classified as the first category are left in the yarn, while events classified as the second category are removed from the yarn.

[0079] In other implementations, the classification system may include more than two categories. A first example of this classification system is one that has the following categories for yarn events: (i) Permitted for all types of further processing; (ii) Only weft yarns used for weaving are not acceptable; (iii) Not acceptable for use in weaving; (iv) Not acceptable for use in knitting; (v) Not acceptable for all types of further processing.

[0080] A second example of a classification system with more than two categories is a classification system for yarn events with the following categories: (i) Acceptable for all colors; (ii) Not acceptable for use on white fabrics / knitwear; (iii) Fabrics / knitwear unacceptable for dyeing; (iv) Not acceptable for all colors.

[0081] Figure 10 The flowchart illustrates a method according to the invention for providing at least one reference dataset to a method or apparatus for classifying events according to the invention. The method can be performed during a specially reserved learning phase and / or during production, for example, during yarn cleaning.

[0082] There must exist a reference event 1001 that will provide a reference dataset.

[0083] In one implementation, the reference event exists in the same test material as other events therein that will be classified later. In a yarn cleaning system, this can be achieved by: examining sufficiently long yarn segments for the reference event during a learning phase prior to yarn cleaning and classifying the reference event as described below; or by continuously using and classifying yarn events discovered during yarn cleaning as reference events. According to the first alternative, reference events below cleaning curve 830 can also be readily considered (see...). Figure 8 This is difficult when the yarn cleaning function is activated. The second alternative has the advantage that the yarn is cleaned starting from the first meter, and the reference database is continuous and complete; however, no reference dataset is available at the start of the cleaning process. Neither alternative is mutually exclusive and they can be combined with each other.

[0084] In another embodiment, the reference event and the event to be classified exist in different test materials. However, these different test materials should have similar characteristics to the test material where the event to be classified is located. According to this embodiment, at least one reference dataset can be provided, for example, in a specially equipped winding station 101 of the yarn winding machine 100 (see...). Figure 1 The reference dataset, provided in this way, can be used to classify yarn events at all other winding stations 101. Alternatively, the reference dataset can be provided by a vendor and made available to the spinning mill.

[0085] Select reference parameter 1002 for the current reference event. Measure at least one reference value for event parameter 1003. Optionally repeat this step for other reference values ​​1004 and other reference parameters 1005. Use the same reference parameter for each reference event.

[0086] Capture an image of the 1006 reference events. This can be done with a camera. Figure 1 A camera 130 is schematically shown at a specific winding position 102, which may be referred to as the "pilot winding position". Camera 130 may be a digital camera with a two-dimensional image sensor. Alternatively, for the moving yarn 110, the camera may only have a line sensor, and successive images from the line sensor can be combined to form a two-dimensional image. Camera 130 may be connected to the measuring head 121 at the pilot winding position 120, or directly to the control unit 123, to transmit images captured by the camera to the control unit 123. Other winding stations 101 do not require cameras. Those winding stations 101 that process the same type of yarn 110 as the pilot winding station 102 may use a reference dataset provided for this yarn type from database 125 to classify yarn events.

[0087] The image is classified at point 1007. Classification at 1007 is performed by the operator. The image is then output to the appropriate output unit 126 (see [link]). Figure 1 The image is displayed to the operator so that the operator can visually evaluate it. The operator evaluates the reference events depicted in the image and classifies the reference events according to their evaluation. They input the classification into the device 120 according to the invention via a suitable input unit 127. Corresponding instructions, such as "Please enter your classification," can be displayed to the operator on the output unit. Thus, classification is performed through guided human-computer interaction.

[0088] Alternative implementations do not require image capture. Operators assess events directly, rather than through images, for example, by observing the events with the naked eye or under a microscope. For this, the relevant coil position must be paused until the event classification is complete.

[0089] Assign the classification 1008 to at least one reference value, and store it together with the reference value 1009 as a reference dataset.

[0090] These steps can be repeated for another reference event 1010. In this way, a database 125 with a reference dataset is constructed.

[0091] Figure 11 Database 125 is shown schematically (see Figure 1Table 1100 of the database contains a reference dataset for the method and apparatus according to the invention. Each row 1111, 1112, ... of Table 1100 contains a data tuple relating to a specific reference dataset. The first column 1121 of Table 1100 may contain a sequence number as the primary key of Table 1100. The second column 1122 contains at least one reference value, for example, a reference signal (see Figure 2(b)). The third column 1123 contains an image of the corresponding reference event. Finally, the fourth column 1124 contains a classification of the corresponding reference events.

[0092] The method according to the invention can be performed as an extension of known yarn cleaning methods. In this case, cleaning limits are specified, for example, according to... Figure 8 The cleanup curve takes the form of 830. Reference events 811 and 812 define exceptions to cleanup based on the cleanup limits and override the cleanup rules given by those limits. In this case, it is advantageous to design the similarity criterion in two stages (i.e., using the "absolute" similarity criterion and the "relative" similarity criterion). The relative similarity criterion has been discussed in detail above. The absolute similarity criterion defines how similar the reference value must be to the current measurement value to be considered in the relative similarity criterion.

[0093] In the example of signals 210 and 220 according to Figure 2, a threshold for the maximum value of the cross-correlation function can be specified for the absolute similarity criterion. The reference signal is only considered for the classification of the event when the maximum value of the cross-correlation function between the reference signal 220 and the measured signal 210 of the event is greater than or equal to the threshold.

[0094] Figure 4 The first graphical representation of the absolute similarity criterion is shown, as a circular sector 420 around the ideal corner point P(1,1), and Figure 8 The second graphical representation is shown as a disk 820 surrounding a reference point 811. If the measured value of an event lies within one of the shaded regions 420 or 820, i.e., sufficiently close to the reference event, then the absolute similarity criterion is satisfied. This applies to, for example... Figure 4 Point 410 in the diagram. If no reference value satisfies the absolute similarity criterion for the current measurement, i.e., if the event is too far from all reference events, the event is evaluated as usual according to the cleanup rules given by the cleanup limits. The cleanup rules given by the cleanup limits can only be overridden if at least one reference event is sufficiently close to the event. This implementation has the advantage that it works even with only a few reference datasets, or even just a single reference dataset.

[0095] On the other hand, the method according to the invention can be performed autonomously without conventional cleaning limits. The yarn is then cleaned solely based on an existing reference dataset. The advantage of this embodiment is that the yarn is cleaned only based on manually performed classifications rather than abstract cleaning limits. The disadvantage is that a large number of reference datasets are required to cover as many yarn event instances as possible.

[0096] It should be understood that the present invention is not limited to the embodiments discussed above. Those skilled in the art will be able to derive other variations upon understanding the present invention, and these variations also fall within the scope of the invention as defined in the independent patent claims.

[0097] List of reference numerals

Claims

1. A computer-implemented method for classifying events (540, 640, 740) in a slender textile test material (110), the method comprising the steps of: a) Measure (901) at least one measured value of at least one event parameter for the events (540, 640, 740); b) Provide (902) at least one reference dataset, wherein each reference dataset contains at least one reference value for the at least one event parameter and a classification of the reference event for the reference event; c) Compare the at least one measurement with the at least one reference value in all reference datasets (904); as well as d) Classify the events (540, 640, 740) according to the classification of the reference events (908), wherein at least one reference value of the reference event is most similar to the at least one measurement value according to a predetermined similarity criterion.

2. The method according to claim 1, wherein The at least one event parameter is the number of events, for example, the deviation of the length-related mass density, cross-sectional area, or reflectance of the test material (110) from the target value, and The at least one reference value and the at least one measurement value are each signals (220, 210) consisting of multiple values ​​of the number of events that vary with length position or time.

3. The method of claim 2, wherein the similarity criterion includes the cross-correlation of the two signals.

4. The method according to claim 1, wherein the at least one event parameter includes the number of events, for example, the length-related mass density, cross-sectional area or reflectance deviation of the test material (110) from the target value, and the event length.

5. The method of claim 4, wherein the similarity criterion includes the distance between a point representing the event and a point representing the reference event (811, 812) in a coordinate system defined by the event length and the event amplitude.

6. The method according to any one of the preceding claims, wherein several different event parameters are considered.

7. The method of claim 6, wherein at least one first event parameter is measured capacitively and at least one second event parameter is measured optically.

8. The method according to any one of the preceding claims, wherein the measured value of the at least one event parameter is measured (901) while the elongated textile test material (110) is moved along its longitudinal direction.

9. The method according to any one of the preceding claims, wherein the at least one reference dataset is provided by measuring (1003) the at least one reference value for each reference event. Measure at least one reference value as described in (1003), The operator visually assesses and classifies the reference events (1007), and The classification is assigned to the at least one reference value (1008) and stored together with the reference value (1009) as a reference dataset.

10. The method of claim 9, wherein the reference event is visually evaluated by capturing (1006) an image (520, 620, 720) of the reference event and presenting the image (520, 620, 720) to the operator for visual evaluation.

11. The method according to any one of the preceding claims, wherein the classification (908) is performed in a classification system having exactly two categories, namely a first category having permissible events and a second category having non-permissible events.

12. A computer-implemented method for cleaning a yarn (110) moving along its longitudinal direction, wherein events (540, 640, 740) in the yarn (110) are classified according to claim 11, and events classified as the second category are removed from the yarn (110).

13. An apparatus (120) for classifying events (540, 640, 740) in a slender textile test material (110), the apparatus comprising: a) A measuring device (121) for measuring at least one measured value of at least one event parameter for the events (540, 640, 740); as well as Computer system (123), which has b) A memory (125) for storing at least one reference dataset, wherein each reference dataset contains at least one reference value for the at least one event parameter and a classification of the reference event for a reference event; c) A processor configured to compare the at least one measurement with the at least one reference value from all reference datasets; as well as d) A processor configured to classify the events (540, 640, 740) according to the classification of the reference events, wherein at least one reference value of the reference event is most similar to the at least one measurement value according to a predetermined similarity criterion.

14. The apparatus (120) according to claim 13, wherein The at least one event parameter is the number of events, for example, the deviation of the length-related mass density, cross-sectional area, or reflectance of the test material from the target value. The at least one reference value and the at least one measured value are each signals (220, 210) consisting of multiple values ​​of the number of events varying with length, position, or time. The measuring device (121) is configured to receive the signal (210).

15. The apparatus (120) of claim 13, wherein the at least one event parameter includes the number of events, for example, the length-dependent mass density, cross-sectional area or reflectance deviation of the test material (110) from the target value, and the event length.

16. The apparatus (120) according to any one of claims 13 to 15, wherein the measuring device (121) is configured to measure several different event parameters.

17. The apparatus (120) according to claim 16, wherein the measuring device (121) is designed for capacitive measurement of at least one first event parameter and for optical measurement of at least one second event parameter.

18. The apparatus (120) according to any one of claims 13 to 17, wherein the processor is configured to perform classification in a classification system having exactly two categories, namely a first category having permissible events and a second category having non-permissible events.

19. A yarn cleaning system for cleaning yarn (110) moving along its longitudinal direction, the yarn cleaning system comprising the apparatus (120) according to claim 18 and a cutting device that can be triggered by the computer system (123) to cut the yarn (110) by classifying an event into the second category.

20. A yarn processing machine, such as a winding machine (100) or a spinning machine, the yarn processing machine having a plurality of yarn processing stations (101, 102), wherein at each yarn processing station, yarn (110) is wound onto a bobbin (112), the yarn processing machine comprising the yarn cleaning system according to claim 19.

21. The yarn processing machine according to claim 20, wherein the yarn processing machine comprises: At least one camera (130) is set at one of the yarn processing stations (102) to capture images (520, 620, 720) of events (540, 640, 740) in the yarn (110); An output unit (126) is connected to the at least one camera (130) to output images (520, 620, 720) captured by the camera (130); as well as An input unit (127) is connected to the processor to input and output the classification of images (520, 620, 720).

22. A computer-implemented method for providing at least one reference dataset for the method according to any one of claims 1 to 12 or the apparatus according to any one of claims 13 to 21, wherein for each reference event in the slender textile test material (110), For each reference event, measure at least one reference value for at least one event parameter (1003). The operator visually assesses and classifies the reference events (1007), and The classification is assigned to the at least one reference value (1008) and stored together with the at least one reference value (1009) as a reference dataset.

23. The method of claim 22, wherein the reference event is visually evaluated by capturing an image (520, 620, 720) of the reference event and presenting the image (520, 620, 720) to the operator for visual evaluation.

24. A computer-readable medium having stored thereon at least one reference dataset provided according to claim 22 or 23.

Images