DE102022004857A1 - Method for determining the coil quality of a collecting coil - Google Patents

Abstract

The invention relates to a method and a device for determining the bobbin quality of a collecting bobbin (14) with synthetic thread (30), wherein a total error (61) is collected during the production of a collecting bobbin (14), wherein the total error (61) is evaluated as a quality-relevant error (63) or a quality-irrelevant error (62), wherein the quality-relevant error (63) is used to determine a quality level (7) of the collecting bobbin (14).

Description

The invention relates to a device for the method for determining the bobbin quality of a collecting bobbin with synthetic thread according to the features of claim 1 and a device for determining the bobbin quality of a collecting bobbin with synthetic thread according to the features of claim 11.

A method for monitoring a large number of processing points for synthetic threads is known, for example, from WO 2019/137835 A1 known.

During the production and treatment of synthetic threads, it is common practice to continuously monitor product and/or process parameters in order to achieve the most stable process control possible and, in particular, the most stable production quality possible for the thread.

Especially in the production of textured threads, monitoring the thread tension on the running threads in the respective processing station has proven to be useful in order to detect process disturbances and/or production fluctuations at an early stage.

In the known method for monitoring the manufacturing process, a thread tension is continuously measured on the running thread at a processing station.

The measurement signals of the thread tension generated are compared with a threshold value of a permissible thread tension in order to identify so-called error graphs and use them for further analyses.

Using appropriate learning algorithms, the respective causes of errors in the manufacturing process are identified from the error graph in order to be eliminated, for example, by an operator.

However, with the current machine learning program it is not possible to determine which types and qualities of errors have been detected and forwarded by the program.

Nevertheless, an initial coil quality assessment is carried out based on the collected errors.

The problem is that even errors that are not relevant to quality are taken into account when evaluating the coil quality.

The object of the invention is therefore to provide a method and a device which make it possible to evaluate the coil quality only on the basis of quality-relevant defects.

With regard to the method, the object is achieved by a method according to the features of claim 1.

The object is achieved with regard to the device with a device having the features of claim 11.

Further developments of the method and the device are claimed in the subclaims.

According to one aspect of the invention, a method is provided for determining the bobbin quality of a collecting bobbin with synthetic thread, wherein a total error is collected during the production of the collecting bobbin, wherein the total error is evaluated as a quality-relevant error or a quality-irrelevant error, wherein the quality-relevant error is used to determine a quality level of the collecting bobbin.

A total error is understood to mean in particular an error that was detected during the production of the synthetic thread. The total error can be reported to a device for producing the synthetic thread, for example to a data analysis unit or a controller of the device for producing the synthetic thread.

A synthetic thread is understood here to be a thread that has been produced by means of a synthetic melt to form a POY or pre-oriented thread made of a plurality of filaments, which is then processed in post-processing steps to form a textured thread that is particularly suitable for processing in the textile industry. The textured thread is also referred to as DTY thread (Drawn Textured Yarn).

This method makes it possible to assess spools according to their actual quality, so that it can be ensured that the synthetic threads collected on the collecting spool are suitable for further processing.

The quality of the threads can be relevant, for example, for a weaving process and/or for subsequent dyeing, whereby a good quality thread has better properties for subsequent processing and the risk of poor end products can be minimized.

According to one embodiment of the method for determining the bobbin quality of a collecting bobbin, the total error is determined from the data determined by a sensor device, in particular a thread tension sensor device.

Currently, it is generally possible to determine thread quality through sensors that monitor the thread during production.

In particular, the measurement of the thread tension with an appropriate measuring device that actively measures the thread tension has become established.

However, there are also other devices that make it possible to determine the thread tension directly or indirectly. For example, this can be determined indirectly via drafting systems and their power consumption.

The required amount of force, such as the torque of a godet or a thread roller guide to set the thread tension, can also be determined.

According to a further embodiment of the method for determining the coil quality of a collecting coil, the determined data of the sensor device are recorded in a graph that defines a tolerance range, wherein the graph is analyzed by means of a first machine learning program for determining total errors in order to detect the total error.

A machine learning program is a software module and/or an AI application that has been programmed to independently solve tasks or independently recognize patterns. This can include detecting and/or reporting possible errors in the production of a synthetic process from certain measurement data. Programming can also include teaching.

The first machine learning program for determining total errors essentially reports the quantity of errors and can also be referred to as the machine learning program for determining error quantities.

To determine a total error from the data of the sensor device, the data is recorded in a time diagram, where the abscissa indicates the time course and the ordinate the respective measured value of the sensor device.

In a graph of the thread tension measurement, the time value is entered on the abscissa and the tension value on the ordinate. This results in a typical graph from which any errors that occur can be analyzed based on the progression and pattern that the graph depicts.

Using a first machine learning program for detecting overall errors, to which AI (artificial intelligence) can be added, the graphs can be analyzed accordingly.

The machine learning program can be programmed for the task using a teach-in. After the teaching, it is possible for the machine learning program to independently detect errors and report them.

According to a preferred embodiment of the method for determining the coil quality of a collecting coil, a first quality level is determined by means of the total error and the first machine learning program, which is assigned to the collecting coil.

For an initial classification of the collecting coil, the total errors determined are collected and assigned with the help of the machine learning program and/or an additional program to the collecting coil that is currently being produced.

A first quality level is an initial assessment of the quality of the collecting coil based on the quantity of total defects detected.

This is stored in a database. This means that an initial spool quality can be determined based on at least the total errors that occurred during the production of the collecting spool and the associated synthetic thread.

However, the method according to the invention allows a more precise determination of the bobbin quality in a further step by assessing the total defects according to their relevance for the thread quality.

According to a special embodiment of the method for determining the coil quality of a collecting coil, the evaluation of the overall defect in a quality-relevant defect or in a quality-irrelevant defect is carried out by means of a second machine learning program for assessing defect qualities.

Evaluating or analyzing the total error with regard to quality relevance allows a better assessment of the spool quality of the associated collecting spool with the thread wound on it.

A quality-relevant defect is a defect that is discovered during the production of the synthetic thread and that reduces the quality of the thread and impairs its further processing. These can be defects that, for example, impair the subsequent dyeing and/or weaving of the thread into a textile.

A quality-irrelevant defect is a defect that is detected during the manufacture of the synthetic thread and does not change the quality of the thread. Such a detected defect can be caused by an error message from a single filament break or an additional measurement with a thread tension measuring device. A quality-irrelevant defect does not affect the quality and processability of the thread.

This has the advantage that collection coils which, based on the initial assessment of the overall defect, fall into a poorer quality level, can fall into a better quality level by reassessing the overall defect into a quality-relevant or quality-irrelevant defect, since only the quality-relevant defects are taken into account.

This also has the advantage that fewer collection spools are assessed incorrectly. This is because the process enables a more precise determination of the quality of the respective collection spool, so that it can be further processed using high-quality processes. This also helps to avoid rejects and waste.

In addition, even collected spools that would originally not be suitable for further processing based solely on the basis of the overall defect can be qualified to a higher quality level through re-evaluation, so that unnecessary waste production can also be avoided.

According to a special embodiment of the method for determining the coil quality of a collecting coil, a maximum target value and a minimum target value for limiting the tolerance range for error detection are determined by means of a third machine learning program for defining tolerance ranges.

The maximum and minimum setpoints can also be determined by predetermined empirical values, which are set, for example, by an experienced operator.

However, in a special embodiment of the method, the target values are monitored and further defined by means of a third machine learning program, whereby it is also possible to train the machine learning program. The target values define the tolerance range with maximum and minimum values, whereby exceeding or falling below the respective target value can indicate an error.

This error can have a significant impact on the quality of the coil. The error is also a warning that the permissible target values have been exceeded or not reached. If the set target values are exceeded or not reached by too much, the production process may even be interrupted.

According to a further embodiment of the method for determining the coil quality of a collecting coil, the first machine learning program for detecting overall errors, the second machine learning program for assessing error qualities and/or the third machine learning program for defining tolerance ranges has a learning process from an operator, wherein the learning process is requested from the operator by the first, second and/or third machine learning program.

For optimal detection of errors and/or optimal monitoring and setting of the setpoints, it is advantageous if the associated machine learning programs are further trained despite the available and recorded data, since these machine learning programs can also recognize patterns and associated errors that cannot be recognized by an operator.

In order for error detection to be accurate and correct, the newly detected errors by the machine learning program should also be confirmed by an experienced operator so that no unnecessary false alarms occur.

According to a further embodiment of the method for determining the coil quality of a collecting coil, the result of the first machine learning program, the second machine learning program and/or the third machine learning program is evaluated by a quality algorithm to determine a second quality level of the collecting coil and stored in a database.

The quality algorithm is another instance by which the quality of the collection coil can be determined, in particular, the quality algorithm plays the role of re-evaluating the first results of detecting the first quality level based on the assessed quality-relevant or quality-irrelevant errors.

The quality algorithm summarizes the results from the defect detection and the defect quality detection and re-evaluates the coil quality based on these results, so that in the best case the coil quality can improve if the majority of the detected defects are not quality-relevant.

According to a particular embodiment of the method for determining the coil quality of a collecting coil, the second machine learning program has an error detection algorithm which reports the quality-relevant error and/or the quality-irrelevant error to the quality algorithm.

In a special embodiment, the second machine learning program for error quality detection not only carries out an evaluation of the total errors, but also reports to the error evaluation of the quality algorithm which types of errors are present.

In simplified form, only the quality-relevant or only the quality-irrelevant errors can be reported to the quality algorithm so that it can evaluate which errors have been reported among the total errors.

The quality algorithm program can then improve or worsen the previously buffered coil quality of the collecting coil, depending on how many of the total defects are quality-relevant or quality-irrelevant.

• - Bereitstellen eines unbearbeiteten Fadens einer Vorlagespule,
• - Bearbeiten des unbearbeiteten Fadens zu einem texturierten Faden,
• - Sammeln von Gesamtfehlern, die bei der Bearbeitung des unbearbeiteten Fadens erfasst werden,
• - Bestimmen der ersten Qualitätsstufe basierend auf den Gesamtfehlern,
• - Sammeln des texturierten Fadens auf einer Sammelspule,
• - Zuordnen der Qualitätsstufe der zugehörigen Sammelspule mit dem texturierten Faden,
• - Bewerten und Markieren der Gesamtfehler in Fehlerqualitäten, die qualitätsrelevante Fehler und qualitätsirrelevante Fehler aufweisen
• - Abziehen der qualitätsirrelevanten Fehler von den Gesamtfehlern zum Erhalten von bereinigten Fehlern,
• - Bestimmten einer zweiten Qualitätsstufe basierend auf den Fehlerqualitäten, und
• - Zuordnen und Speichern der zweiten Qualitätsstufe der zugehörigen Sammelspule.

According to a preferred embodiment of the method for determining the coil quality of a collecting coil, the method comprises the following steps

• - Providing an unprocessed thread from a supply spool,
• - Processing the unprocessed thread into a textured thread,
• - Collecting total errors detected during processing of the unprocessed thread,
• - Determining the first quality level based on the total errors,
• - Collecting the textured thread on a collecting spool,
• - Assigning the quality level of the corresponding collection spool with the textured thread,
• - Evaluating and marking the total defects in defect qualities that contain quality-relevant defects and quality-irrelevant defects
• - Subtracting the quality-irrelevant errors from the total errors to obtain corrected errors,
• - Determining a second quality level based on the defect qualities, and
• - Assigning and saving the second quality level of the associated collection coil.

According to an embodiment of the above method, the total defects detected can be classified into quality-relevant and quality-irrelevant defects by means of the method steps described above and a new quality assessment of the collecting spool can be carried out, whereby the defects are taken into account according to their relevance for the quality of the thread produced.

This can result in a collection coil that was initially assigned to a poorer quality level being upgraded to a better quality level.

According to one aspect of the invention, a device is provided for determining the bobbin quality of a collecting bobbin with synthetic thread, in particular for the method according to the embodiments listed above, comprising a sensor device for detecting data of the synthetic thread, a data analysis unit with an error detection unit and error evaluation unit, which cooperates with a control of an operating station for producing the synthetic thread, wherein the error evaluation unit cooperates with a quality determination unit for determining a quality level, and a database for storing the quality level of the associated collecting bobbin.

With the device it is possible to carry out an accurate quality assessment of the produced collecting coil, whereby in particular the devices mentioned above are used so that an accurate quality assessment of the collecting coil can be carried out in a targeted manner in order to be able to evaluate high-quality collecting coils.

According to an embodiment of the device for determining the bobbin quality of a collecting bobbin with synthetic threads, the quality determination unit is assigned a tolerance range determination unit which transmits current threshold values of the tolerance range to the quality determination unit.

An advantageous additional device to improve the device is a tolerance range detection unit that regularly checks and monitors the specified threshold values for the tolerance range so that accurate error detection is possible.

The invention relates to a method and a device with which a re-evaluation of the detected errors takes place in which the collected total errors are sorted according to their effect on the thread quality. In this case, at least the harmless errors that have no effect on the thread quality are evaluated, marked and/or identified so that they can be distinguished from the errors that can have a negative effect on the thread quality. This is also possible for the quality-relevant errors.

The harmless errors can be detected using pattern recognition of previously determined sensor data. The pattern recognition uses a machine algorithm, similar to an AI, which can be trained by teaching harmless errors. This is also possible for quality-relevant errors.

This makes it possible to carry out a new assessment of the quality of collecting spools if the first quality assessment is based on total defects already detected and a second quality assessment is carried out by weighting the total defects detected at least according to quality-relevant and quality-irrelevant for the thread quality.

The method according to the invention for determining the coil quality of a collecting coil and the associated device for determining the coil quality of a collecting coil are explained in more detail below with reference to the figures and with explanation of further exemplary embodiments with reference to the attached figures. This results in further embodiments and advantages of the method and the device.

• 1 schematisch ein Ausführungsbeispiel einer Bearbeitungsstelle einer Textilmaschine zur Herstellung eines synthetischen gekräuselten Fadens mit einer zugehörigen Datenanalyseeinheit,
• 2 schematisch ein erstes Maschinenlernprogramm zum Definieren und Erkennen von Gesamtfehlern mittels denen eine erste Qualitätsstufe der Sammelspule und den darauf gesammelten synthetischen Faden definiert werden kann,
• 3 schematisch ein drittes Maschinenlernprogramm zum Definieren von Toleranzbereichen, die durch zu bestimmenden maximalen und minimalen Schwellwerte begrenzt sind,
• 4 ein wird drittes Maschinenlernprogramm zum Erkennen von Fehlerqualitäten, insbesondere Fehlerqualitäten, die die Spulenqualität beeinflussen und von Fehlern, die die Spulenqualität nicht beeinflussen, und
• 5 eine Zusammenschau der Datenanalyseeinheit mit deren zugehörigen ersten, zweiten und dritten Maschinenlernprogrammen die zur Beurteilung der Spulenqualität mittels eines Qualitätsalgorithmus herangezogen werden können.

The drawings show:

• 1 schematically an embodiment of a processing station of a textile machine for producing a synthetic crimped thread with an associated data analysis unit,
• 2 schematically a first machine learning program for defining and detecting total errors by means of which a first quality level of the collecting spool and the synthetic thread collected on it can be defined,
• 3 schematically a third machine learning program for defining tolerance ranges limited by maximum and minimum threshold values to be determined,
• 4 a third machine learning program for detecting defect qualities, in particular defect qualities that affect the coil quality and defects that do not affect the coil quality, and
• 5 an overview of the data analysis unit with its associated first, second and third machine learning programs which can be used to assess the coil quality using a quality algorithm.

In 1 is a schematic representation of an embodiment of a processing station 1 of a textile machine, in this case a texturing machine for producing a textured thread 30 (DTY = Drawn Textured Yarn - crimped synthetic thread) from an untextured thread 3, which is also referred to as POY (Pre Oriented Yarn). Such texturing machines have a large number of such processing stations 1, which are constructed identically. It is therefore usual for more than 100 processing stations to be arranged next to one another along a machine's long side. At this point, only one of the processing stations 1 of a textile machine is described in more detail.

In 1 the processing station 1 and a winding station 2 of the textile machine are shown schematically. The processing station 1 has a creel 4 in which a supply spool 5 and a reserve spool 6 are held. The supply spool 5 supplies a thread 3 which is transferred to the processing station 1 for stretching and texturing. The supply spool 5 is also referred to as a so-called POY spool because it was previously produced in a melt spinning process and wound with a freshly spun POY thread. A thread end of the supply spool 5 is connected to a thread start of the reserve spool 6 by a thread knot. In this way, a continuous draw-off of the thread 3 is achieved after the supply spool 5 has been completely unwound. The thread end of the reserve spool 6 is then connected to the thread start of a new supply spool 5.

The thread 3 is drawn off from the feed spool 5 by a first feed mechanism 7.1. The feed mechanism 7.1 is driven by a drive 8. In this embodiment, the feed mechanism 7.1 is formed by a driven godet and a freely rotating roller, around which the thread is wrapped several times. A heating device 9, a cooling device 10 and a texturing unit 11 are arranged downstream of the feed mechanism 7.1. The texturing unit 11 is preferably designed as a friction twister in order to Thread to create a false twist, which results in a crimping of the individual filaments of the thread. To stretch the thread, a second delivery mechanism 7.2 is arranged downstream of the texturing unit 11, which is driven by the drive 8. The delivery mechanism 7.2 is identical in structure to the first delivery mechanism 7.1, whereby the second delivery mechanism 7.2 is operated at a higher peripheral speed to stretch the thread. Within the processing station 1, the synthetic thread 3 is thus textured and stretched at the same time.

After the yarn 3 has been crimped, it is guided to the winding station 2 by a third feed mechanism 7.3. The feed mechanism 7.3 is driven by the drive 8. The feed mechanism 7.3 is designed as a so-called clamping feed mechanism, which has a driven shaft and a pressure roller. The yarn 3 is guided in a clamping gap on the circumference of the shaft.

The winding station 2 has a bobbin holder 13 which carries a collecting bobbin 14. The collecting bobbin holder 13 is designed to be pivotable and can be operated manually or automatically to replace the collecting bobbins 14. The collecting bobbin holder 13 is assigned a drive roller 15 which is driven by a roller drive 15.1. To lay the thread on the circumference of the collecting bobbin 14, the winding station 2 is assigned a traversing device 12 which has a drivable traversing thread guide. For this purpose, the traversing thread guide is driven in an oscillating manner via the traversing drive 12.1. The traversing drive 12.1 and the roller drive 15.1 of the winding station 2 are designed as individual drives and are connected to a machine control unit 16. Likewise, the drives 8 of the delivery units 7.1, 7.2 and 7.3 as well as the texturing drive 11.1 of the texturing unit 11 of the processing station 1 are designed as individual drives and coupled to the machine control unit 16.

To monitor the manufacturing process, a thread tension on the running thread is continuously recorded in each of the processing stations 1 of a textile machine. For this purpose, a sensor device 17 is arranged in the thread path in front of the delivery mechanism 7.2 at the exemplary processing station 1.

The sensor device 17 has a thread tension sensor 17.1 and a measurement signal sensor 17.2. The sensor device 17 is connected to a data analysis unit 18. The data analysis unit 18 is coupled to a transmitter 19, which is connected to suitable receiving systems using wired or wireless transmission technologies in order to transmit information and signals. The data analysis unit 18 has several software modules or machine learning programs 35, 36, 37 in order to analyze the measurement signals of the sensor device 17 and the data obtained therefrom in order to monitor the manufacturing process.

The data recorded and evaluated by the data analysis unit 18 can be forwarded to a database 66 and/or the machine control unit 16 using the transmitter 19. The transmitter 19 can act as a transmitter and receiver and transmit the data wirelessly and/or via cable.

At the start of a process that requires a product change in the supply spools or after a certain running time of the process units, it is usual to check the thread tension state by means of a manual measurement. For this purpose, an operator uses a thread tension measuring device 25, which the operator holds in the thread path of the processing point 1 to be checked.

In the 1 the thread tension measuring device 25 operated by the operator is shown in dashed lines. The thread tension measuring device 25 is preferably held in the thread path in front of or behind the sensor device 17. The measured value of the thread tension displayed by the thread tension measuring device 25 is checked by the operator to see whether it lies within a predefined tolerance range of the manufacturing process. However, the intervention of the thread tension device 25 generates an event on the sensor device 17 that leads to a deviation from the standard state of the thread tension in the manufacturing process. This event is fed from the thread tension sensor 17.1 and the measurement signal sensor 17.2 to the data analysis unit 18.

Within the data analysis unit 18, a thread tension analysis module is provided, which continuously records the measuring signals of the measuring signal sensor 17.2 and generates so-called error graphs.

The graphs 351 define an abnormal measurement signal curve of the thread tension. Such graphs 351 are always generated by data technology when the thread tension measured online in the manufacturing process leaves the range 352 defined by the specified limit values OG, OG. If the thread tension of the thread moves significantly outside the defined range 352, this has an effect on the quality of the thread and thus on the quality of the manufactured product. In this way, every event that changes the thread tension is recorded and generated as an error graph.

In 2 The error graph 351 is shown schematically, which is assigned to a manual thread tension measurement as the cause of the error during the analysis. The determination and analysis of the error graphs are based on the EP 3 737 943 A1 and are described in more detail there.

There are a large number of error graphs 351 that are analyzed using machine learning programs to identify the source of the error.

In the 2 The error graph 351 shown also shows a typical measurement signal curve of the thread tension T, which occurs when a thread tension measuring device 25 is used for manual measurement of the thread tension. For this purpose, the thread tension T is plotted on the ordinate and the time t on the abscissa.

The intervention of the thread tension measuring device 25 in the thread path produces an increase in the thread tension T, which remains essentially constant during the manual measurement of the thread tension. In this respect, the manual thread tension measurement can be clearly assigned to the error graph with the help of the machine learning program.

As further stated 2 As can be seen, the error graph 351 has threshold values which indicate a maximum value OG and a minimum value UG, which also define a tolerance range 352.

If the measured thread tension remains within the tolerance range 352, no total error 61 is displayed. However, if the thread tension is outside the tolerance range 352, a total error 61 is generated, which is then collected by means of the machine learning program 35 and assigned to the associated collecting bobbin 14, so that a first quality level 40, which is in the 2 indicated with a, b or c.

The quality level 40 with a means that the collecting coil is of very high quality and is very suitable for further processing.

For quality level 40 with the designation b, the quality is medium-good and the collecting coil 14 is not suitable for further processing for every product and for quality level 40 with c, further processing of the collecting coil 14 is only possible to a limited extent or not at all.

However, since the total errors 61 can also be the result of errors that are not relevant to quality, such as the thread tension measurement mentioned above, which can also result in the threshold values of the tolerance range 352 being exceeded or not reached and a corresponding error evaluation, according to the invention a second machine learning program 36 is provided for assessing error qualities, with which the quality relevance of a total error 61 can be analyzed.

First, reference is made to 3 taken, which shows a third machine learning program 37, which has a tolerance range determination unit 70 with which the threshold values of the tolerance range for the graph 351 can be defined or learned.

This can be done, for example, via a teach-in 64 and/or the operator specifies the threshold values OG and UG based on experience values of the operating position 1.

In 4 the second machine learning program 36 is shown for detecting quality-relevant errors 63, which are used for the assessment of the collecting coil 14 in a second quality level 200.

In a first step, a total error 61 that has been detected by means of the first machine learning program 35 is fed to the second machine learning program 36. This carries out an error evaluation 60 with a corresponding algorithm and then evaluates the total error 61 accordingly as a quality-irrelevant error 62 or as a quality-relevant error 63.

Intermediate stages can also be incorporated here which further weight the relevance or irrelevance of the error accordingly, e.g. by numerical values from 1 to 10, where 10 is a very high quality-relevant error and 1 is a rather low one, with 0 meaning the error has no effect at all on the quality of the thread of the collecting spool 14.

To perfect the error evaluation unit 60, the knowledge of an operator can also be incorporated by means of a teach-in 64.

A teach-in or training 64 refers to a method for programming a machine learning program. The programming is supported by an operator who contributes his experience to the error messages of the machine learning program.

The error patterns identified in this way are stored in the control of the respective machine learning program 35, 36. This step can be repeated regularly as required. to better control and/or verify the machine learning program's responses.

The program sequence can consist of the machine learning program displaying new or unknown errors to the operator, possibly with a suggested solution or an assessment, and the operator confirming and/or modifying these suggestions if they are correct. In addition, the error evaluation unit 60 can send a warning signal or alarm signal to the operator if a corresponding additional evaluation by the operator or training of the operator is necessary, for example if new, unknown errors occur.

In addition, the error evaluation unit 60 can interact with a quality determination unit 100 and communicate its quality assessment of the respective total errors 61 into quality-relevant errors or quality-irrelevant errors 63, so that the quality determination unit 100 can carry out a new evaluation of the collecting coil 14 with its quality algorithm 101 and can assign it a second quality level 200, which then takes priority for the further life of the collecting coil 14.

In addition, the error evaluation unit 60 is connected to a database 66 in which the corresponding evaluations or the recorded total errors 61 and their assessment as quality-relevant errors 63 or as quality-irrelevant errors 62 are stored, whereby corresponding error levels 65 - indicated with I, II or III - can also be assigned to the total error 61 or the quality-relevant error 63.

This is always stored in the database 66 linked to the associated collecting coil 14 so that it can be retrieved over the entire service life of the collecting coil 14 and beyond.

In 5 a summary of the possible embodiments of a data analysis unit 18 for the inventive method or for the inventive device for determining the coil quality of a collecting coil 14 is shown.

In the embodiment shown, the data analysis unit 18 can have three machine learning programs 35, 36 and 37.

Each of the machine learning programs 35, 36, 37 has a different function. Their results are finally checked with a quality determination unit 100, which has a quality algorithm 101. With the help of the quality determination unit 100 and its quality algorithm 101, a second quality level 200 is then determined, which is then assigned to the respective collecting coil 14.

This can, for example, result in an improvement of the previously determined first quality level 40 of the collecting coil 14 based on the results or definitions of the second machine learning program 36 and third machine learning program 37.

The first machine learning program 35 has an error detection unit 50 that can detect total errors 61 using the previously presented graphs 351. Based on the total errors 61, the collecting spool 14 is assigned a first quality level 40. However, these total errors 61 can also have errors that are not relevant to quality, such as a thread tension measuring device measurement that can exceed the target values of the tolerance range 352 but has no effect on the quality of the thread 30.

These non-quality-relevant defects may also include filament breakages, which may lead to a change in the thread tension but do not affect the quality processing or the processability of the collecting spool 14 and the synthetic threads wound on it.

An assessment of the total errors 61 takes place with the second machine learning program 36, which has an error evaluation unit 60 for detecting and defining quality-relevant errors 62 and quality-irrelevant errors 63.

A teaching of the error evaluation unit 60 as well as a teaching of the error detection unit 50 can be carried out, for example, according to a teach-in method 64, wherein an operator at least confirms the corresponding error indications of the respective unit.

In addition, the data analysis unit 18 can have a third machine learning program 37 which defines at least the tolerance ranges of the error detection.

The third machine learning program 37 has a tolerance range determination unit 70, which determines the threshold values UG, OG of the error detection graphs 351 or adjusts them accordingly in corresponding analyses.

The results of the first, second and/or third machine learning program 35,36,37 are then fed to the quality determination unit 100, which then uses a quality algorithm 101 to re-evaluate the first quality level based on the results of the tolerance range determination unit 70 and the error evaluation unit 60 and also the error detection unit 50. A new second quality level 200 is then fed to the collecting coil 14 and this is stored in the database 66 in connection with the associated collecting coil 14.

In a further embodiment of the method for determining the coil quality, the data analysis unit 18 has either the first and second machine learning programs 35, 36 or the first and third machine learning programs 35, 37.

In another embodiment, the second machine learning program 36 may be programmed solely for error detection and error quality classification and applied in the data analysis unit 18.

• Bereitstellen eines unbearbeiteten Fadens 3 von einer Vorlagenspule 5, wie die die beispielsweise in 1 gezeigt ist.

According to one embodiment of the method for determining the coil quality of a collecting coil, the method may comprise the following steps:

• Providing an unprocessed thread 3 from a supply spool 5, such as the one shown in 1 is shown.

The unprocessed thread 3 can be provided to a gate 4.

In order to ensure unrestricted processing of the thread 3 to be processed, a supply spool 5 and a reserve spool 6 are provided, each with unprocessed thread 3. The unprocessed thread 3 of the supply spool 6 and the reserve spool 5 are each linked to one another, as can be seen from 1 visible.

The unprocessed thread 3 is processed into a textured thread 30.

Preferably, the unprocessed thread 3, as shown in 1 can be seen, via delivery systems 7.1, 7.2 and 7.3 to the heating device 9 and the cooling device 10. In addition, the unprocessed thread 3 is textured by means of the texturing unit 11, so that a textured thread 30 is formed. The textured thread 3 is collected on the collecting spool 40.

During the production of the textured thread 3, detected total errors 61 are collected, which are detected during the processing of the unprocessed thread 3. The detection of the total errors 61 can be carried out, for example, by means of the sensor device 17.

An assessment of the total errors 61 is then carried out using the data analysis unit 18 and its machine learning programs 35, 36 and 37.

Using the first machine learning program 35 for detecting error quantities, total errors 61 and their frequency can be determined. Based on the frequency, a first quality level 40 can be determined in which only the number of total errors 61 is taken into account. The first quality level 40 is assigned to the associated collecting spool 14 with the textured thread 30.

By means of the second machine learning program 36 for detecting and assessing error qualities, the total errors 61 are evaluated and marked into quality-relevant errors 63 and quality-irrelevant errors 62.

By means of the quality determination unit 100 with a quality algorithm 101, a second quality level 200 of the collecting coil can be determined, for example by subtracting the quality-irrelevant errors 62 from the total errors 61 to obtain corrected errors, so that the second quality level 200 can be determined on the basis of the corrected errors.

The second quality level 200 of the associated collection coil 14 is also assigned and stored, for example in the database 66.

As in the 1 and 5 shown schematically, according to one embodiment of the device for determining the bobbin quality of a collecting bobbin 4, a sensor device 19 for recording data of the synthetic thread 30, a data analysis unit 18 with an error detection unit and error evaluation unit 60, which interacts with a controller 16 of an operating station 1 for producing the synthetic thread 30. The error evaluation unit 60 interacts with a quality determination unit 100 for determining a quality level 200. The recorded data and results are stored and preserved in a database 66 for storing the quality level 200 of the associated collecting bobbin 14.

In addition, the quality determination unit 200 is assigned a tolerance range determination unit 70, which transmits updated threshold values OG, UG of the tolerance range to the quality determination unit.

List of reference symbols:

1

Operating location / position

3

untextured thread, pre-oriented yarn, POY

4

gate

5

Feed spool

6

Reserve coil

7.1, 7.2, 7.3

first delivery plant, second delivery plant, third delivery plant

8th

drive

9

Heating device

10

Cooling device

11

Texturing unit

11.1

Texturing drive

12

Traversing device

13

Collecting spool holder

14

Collecting coil

15

Drive roller

15.1

Roller drive

16

Machine control unit

17

Sensor device

17.1

Thread tension sensor

17.1

Measuring signal sensor

18

Data analysis unit,

19

Transmitter

25

Thread tension meter

30

textured thread, DTY, synthetic thread

35

first machine learning program

36

second machine learning program

37

third machine learning program

351

graph

352

Tolerance range

40

first quality level

50

Error detection unit

60

Error evaluation unit

61

Total error

62

quality-irrelevant error

63

quality-related error

64

Training, Teach-In

65

Error levels

66

Database

70

Tolerance range determination unit

100

Quality determination unit

101

Quality algorithm

200

second quality level

Upper floor, basement

Threshold

T

Thread tension

QUOTES INCLUDED IN THE DESCRIPTION

This list of documents listed by the applicant was generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA accepts no liability for any errors or omissions.

Cited patent literature

• WO 2019/137835 A1 [0002]
• EP 3737943 A1 [0082]

Claims (12)

Method for determining the bobbin quality of a collecting bobbin (14) with synthetic thread (30), wherein a total error (61) is collected during the production of the collecting bobbin (14), wherein the total error (61) is evaluated as a quality-relevant error (63) or a quality-irrelevant error (62), wherein the quality-relevant error (63) is used to determine a quality level (200) of the collecting bobbin (14). Method for determining the coil quality of a collecting coil (14) according to Claim 1 characterized in that the total error (61) is determined from the data determined by a sensor device (17), in particular a thread tension sensor device (17.1). Method for determining the coil quality of a collecting coil (14) according to at least one of the preceding claims, characterized in that the determined data of the sensor device (17) are recorded in a graph (351) which defines a tolerance range (352), wherein the graph (352) is analyzed by means of a first machine learning program (35) for determining total errors in order to recognize the total error (61). Method for determining the coil quality of a collecting coil (14) according to at least one of the preceding claims, characterized in that a first quality level (40) is determined by means of the total error (61) and the first machine learning program (35), which is assigned to the collecting coil (14). Method for determining the coil quality of a collecting coil (14) according to at least one of the preceding claims, characterized in that the evaluation of the overall error (61) into a quality-relevant error (63) or into a quality-irrelevant error (62) is carried out by means of a second machine learning program (36) for assessing error qualities. Method for determining the coil quality of a collecting coil (14) according to at least one of the preceding claims, characterized in that the definition of the tolerance range (352) for the error detection is determined by means of a third machine learning program (37) for defining tolerance ranges. Method for determining the coil quality of a collecting coil (14) according to at least one of the preceding claims, characterized in that the first machine learning program (35) for detecting total errors, the second machine learning program (36) for assessing error qualities and/or the third machine learning program (37) for defining tolerance ranges has a learning process (64) from an operator, wherein the learning process (64) is requested from the operator by the first, second and/or third machine learning program. Method for determining the coil quality of a collecting coil (14) according to at least one of the preceding claims, characterized in that the result of the first machine learning program (35), the second machine learning program (36) and/or the third machine learning program (37) is evaluated by a quality algorithm (101) for determining a second quality level (200) of the collecting coil (14) and is stored in a database (66). Method for determining the coil quality of a collecting coil (14) according to at least one of the preceding claims, characterized in that the second machine learning program (37) has an error detection algorithm (60) which reports the quality-relevant error (63) and/or the quality-irrelevant error (62) to the quality algorithm (101). Method for determining the bobbin quality of a collecting bobbin (14) according to at least one of the preceding claims, characterized by - providing an unprocessed thread (3) from a feed bobbin (5), - processing the unprocessed thread (3) to form a textured thread (30), - collecting total errors (61) that are detected during processing of the unprocessed thread (3), - determining the first quality level (40) based on the total errors, - collecting the textured thread (30) on a collecting bobbin (14) - assigning the quality level (40) of the associated collecting bobbin (14) to the textured thread (30), - evaluating and marking the total errors (61) in error qualities that have quality-relevant errors (63) and quality-irrelevant errors (62), - subtracting the quality-irrelevant errors (62) from the total errors (61) to obtain corrected errors, - determining a second quality level (200) based on the corrected errors, and - assigning and storing the second quality level (200) from the associated collecting coil (14). Device for determining the bobbin quality of a collecting bobbin (14) with synthetic thread (30), in particular for the method according to the claims, comprising a sensor device (19) for detecting data of the synthetic thread (30), a data analysis unit (18) with an error detection unit and error evaluation unit (60) which cooperates with a controller (16) of an operating station (1) for producing the synthetic thread (30), wherein the error evaluation unit (60) cooperates with a quality determination unit (100) for determining a quality level (200), and a database (66) for storing the quality level (200) of the associated collecting spool (14). Device for determining the bobbin quality of a collecting bobbin (14) with synthetic thread (30) according to Claim 11 , characterized in that the quality determination unit (200) is assigned a tolerance range determination unit (70) which transmits updated threshold values (OG, UG) of the tolerance range to the quality determination unit.

Images