阅读说明：本技术 用于监视变形过程的方法和设备 (Method and device for monitoring a deformation process ) 是由 J·胡特马赫尔 于 2019-01-02 设计创作，主要内容包括：本发明涉及用于监视用于制造卷曲纱线的变形过程的一种方法和一种设备。在此在变形的纱线上连续地测量纱线张力并且至少在一个时间间隔内连续地测得纱线张力测量信号并进行分析。在这种情况下,根据本发明为了提前诊断多个干扰源之一利用机器自学程序对一个在时间间隔内出现的测量信号序列进行分析。本发明的设备为此具有一个诊断单元,该诊断单元与纱线张力测量装置如此共同作用,使得能够利用机器自学程序对纱线张力测量信号进行分析以辨别出多个干扰源之一。(The invention relates to a method and a device for monitoring a texturing process for producing crimped thread. The yarn tension is continuously measured on the textured yarn and the yarn tension measurement signal is continuously measured and evaluated at least over a time interval. In this case, according to the invention, a sequence of measurement signals occurring within a time interval is analyzed by means of a machine self-learning program in order to diagnose one of a plurality of interference sources in advance. The inventive device has a diagnostic unit which interacts with the thread tension measuring device in such a way that the thread tension measuring signal can be analyzed by a machine-specific program in order to detect one of a plurality of interference sources.)

1. A method for monitoring a texturing process for producing a crimped yarn, in which method the yarn tension is continuously measured on the textured yarn and in which method a yarn tension measurement signal is continuously measured and evaluated at least over a time interval, characterized in that: the sequence of yarn tension measurement signals occurring during this time interval is analyzed by a machine-self-learning program in order to diagnose one of the interference sources in advance.

2. The method of claim 1, wherein: the yarn tension measurement signal sequence was measured as an analysis chart and analyzed.

3. The method of claim 1, wherein: when a threshold value for the yarn tension is exceeded, a yarn tension measurement signal sequence is detected as a defect map and evaluated.

4. The method of any of claims 1 to 3, wherein: the analysis of the yarn tension measurement signal is performed by at least one machine learning algorithm of a machine self-learning program.

5. The method of claim 4, wherein: one of the sources of interference is identified by a machine learning algorithm from the analyzed sequence of measurement signals or from the analyzed analysis map or from the analyzed defect map.

6. The method of claim 5, wherein: the analysis map and/or the fault map are assigned a plurality of interference maps, wherein each of the interference sources is determined by one of the interference maps.

7. The method of any of the preceding claims, wherein: the machine learning algorithm has reached an end-of-learning operating condition after a learning phase.

8. The method of claim 7, wherein: in the presence of unknown sources of interference, the machine learning algorithm is replaced for external learning purposes.

9. The method of any one of claims 1 to 8, wherein: operational errors, incorrect adjustment of the processing units, material defects, wear of the thread guide elements and/or thread knots are one of the sources of interference.

10. The method of any one of claims 1 to 9, wherein: after the identification of the source of interference or after configuration to one of the interference maps, a control command for the change process is triggered.

11. An apparatus for monitoring a texturing process for producing crimped yarn, comprising a yarn tension measuring device (17) for measuring the yarn tension at a measuring point and a data analysis device (18) for analyzing a measurement signal of the yarn tension measuring device (17), characterized in that: the data analysis device is formed by a diagnosis unit (18) by means of which the measurement signals of the yarn tension (T) can be analyzed by means of a machine-based program in order to identify the source of disturbance.

12. The apparatus of claim 11, wherein: the diagnostic unit (18) has a programmable learning processor (20) for executing a machine self-learning program.

13. The apparatus of claim 12, wherein: the learning processor (20) is optionally coupled for learning purposes with an input unit (22) by means of which one or more determined interference patterns of the yarn tension can be read.

14. The apparatus of claim 12 or 13, wherein: the learning processor (20) is coupled to an output unit (23), by means of which the identification of the source of interference and/or the assignment to one of the interference maps can be visualized.

15. The apparatus of any one of claims 112 to 14, wherein: the learning processor (20) has a neural network for performing a machine self-learning procedure.

16. The apparatus of any of claims 12 to 15, wherein: the learning processor (20) is spatially separated from the input unit (22) and the output unit (23).

17. The apparatus of any of claims 11 to 16, wherein: the diagnostic unit (18) is connected to a machine control unit (16), by means of which control commands for changing the process can be executed.

Technical Field

The invention relates to a method for monitoring a texturing process for producing a crimped yarn, as claimed in the preamble of claim 1, and to a device for monitoring a texturing process, as claimed in the preamble of claim 11.

Background

A method of this type and a device of this type for monitoring a texturing process for producing crimped thread are known, for example, from DE 19614027 a 1.

The production process is usually continuously monitored during the texturing of the synthetic thread in order to obtain a possible, stable process control and in particular to maintain as stable a product quality of the crimped thread as possible. In this case, monitoring of the yarn tension on the moving yarn has proven effective in identifying process disturbances and/or product fluctuations. For this purpose, the yarn tension on the yarn is continuously measured in known methods and known devices for monitoring the texturing process. The yarn tension measurement signal generated at the same time is compared with a threshold value for the permissible yarn tension. In this case, the signal curve of the yarn tension measurement signal measured continuously over a time interval shows possible tolerance deviations. In this case, different signal profiles of the measurement signal can be determined depending on the respective disturbance in the process. In this way, experienced operators can use the measured signal profile of the yarn tension to identify possible sources of disturbance to the texturing process.

However, the known method and the known device for monitoring a deformation process have significant disadvantages: only the measured signal curve exceeding the threshold value of the yarn tension is used for the analysis. This already results in unacceptable product limits which lead to defects in the yarn quality. In addition, the known method and the known device have further disadvantages: the identification of possible sources of interference depends only on the experience of the respective operator.

Disclosure of Invention

The object of the invention is therefore to further develop a method of this type and a device of this type for monitoring a texturing process for producing crimped yarn in such a way that improved process control for producing a stable yarn quality can be achieved.

A further object of the invention is to provide a method of this type and a device of this type for monitoring deformation processes, with which process disturbances can be identified as early as possible and can be eliminated quickly and specifically.

This object is achieved according to the invention by a method having the features of claim 1 and by an apparatus having the features of claim 9.

Advantageous developments of the invention are defined by the features and feature combinations of the respective dependent claims.

The invention is based on the following recognition: the measurement signal sequences occurring within a time interval have certain characteristics which are represented in the signal profile and can each be used as an indication for process control. The sequence of yarn tension measurement signals occurring in this time interval is therefore analyzed according to the invention using a machine-self-learning program in order to diagnose one of the interference sources in advance. This enables the previously determined changes in the yarn tension measurement values occurring in time sequence to be recognized in advance on the basis of a known source of disturbance. In the case of a yarn tension measurement that does not exceed a threshold value beforehand, it is therefore possible to identify in advance a measurement signal change that is representative for a source of disturbance. The correlation of the measurement signals with each other based on a source of interference is independent of the threshold value of the yarn tension. This allows the analysis of the variations in the measured signal to derive a characteristic signature for identifying the source of the interference. The machine self-learning procedure makes possible a rapid and complex analysis of a plurality of measurement signals, so that their correlation with one another (in particular over a time interval) is analyzed rapidly and a source of interference can therefore be reliably and rapidly identified.

Since the change in the measurement signal based on an interferer causes a unique, specific, temporal change depending on the interferer, a method variant is particularly advantageous in order to obtain the best possible unambiguous discrimination of the interferer. The yarn tension measuring signal sequence is then preferably determined and evaluated as an evaluation chart.

In order to be able to distinguish, in particular, an admissible change in yarn tension from an inadmissible change in yarn tension which exceeds a predetermined threshold value, the method variant is particularly advantageous in which the time sequence of the yarn tension measurement signal is measured as a defect map and evaluated when the threshold value of the yarn tension is exceeded. The threshold value can also comprise a tolerance range, for example by an upper limit value and a lower limit value of the yarn tension.

The method variant, in which the analysis of the yarn tension measurement signal is carried out by at least one machine learning algorithm of a machine self-learning program, has proven particularly effective on the basis of a plurality of measurement signals which are continuously awaiting analysis. This enables the use of artificial intelligence for structured analysis even in the case of large data volumes and makes it possible to identify the source of interference within the shortest analysis time. For this purpose, however, the machine learning algorithm is first required to refer to the already determined basic data for learning purposes. For this purpose, for example, at the beginning of the process, the analyzed analysis map and the analyzed fault map are transferred to the machine learning algorithm for learning purposes. After a learning phase there is a possibility: the machine learning algorithm automatically carries out a clear identification of the respective interference source by analyzing the measurement signal or the analysis or fault map.

In order to facilitate discrimination in the case of a plurality of interference sources, provision is additionally made for: the analysis map and/or the fault map are assigned a plurality of interference maps, wherein each of the interference sources is determined by one of the interference maps. Such a disturbance map is used for learning of a machine self-learning program, so that the disturbance source can be identified in advance when the analysis map or the fault map is analyzed.

According to a particularly advantageous method variant, the machine learning algorithm reaches the learning-end operating state after a learning phase. The transmission of the interference pattern can be omitted and the machine self-learning program can analyze the measured defect pattern and, if necessary, assign it to a known interference source.

In the case of an unidentifiable unknown source of interference, the machine learning algorithm may be replaced for external learning purposes, so that production can continue with the newly learned machine learning algorithm. In this regard, diagnostic systems can continue to expand to include new, yet unknown, sources of interference.

In this case, malfunctions or incorrect adjustment of the processing assembly, or material defects or wear of the thread guide element or thread knots or other product defects can be identified as sources of interference. The early diagnosis of possible sources of disturbance already before the yarn tension threshold is reached is therefore particularly suitable for performing maintenance of the processing assembly, for example, without rain or without work. This makes it possible, for example, to suppress wear phenomena of the thread guide element in advance.

However, in order to automate the corresponding deformation process, there are also possibilities: after the identification of the source of interference or after configuration to one of the interference maps, a control command for the change process is triggered. The change process may be, for example, an early change of the bobbin to avoid trapping knots. However, it is alternatively also possible to send certain operating instructions to the operator by sending a signal.

The device according to the invention for monitoring a texturing process for producing crimped thread achieves the stated object by: the data analysis device is formed by a diagnostic unit, by means of which the measurement signal of the thread tension can be analyzed by means of a machine self-learning program in order to identify one of the plurality of interference sources. This enables the measurement signals of the yarn tension occurring in the time interval during the deformation to be used directly for the diagnosis of the source of interference.

The diagnostic unit has at least one programmable learning processor for executing a machine self-learning program. In this case, the learning processor can be coupled directly to the yarn tension measuring device.

In order to optimize the machine self-learning program and to increase the diagnostic reliability, provision is additionally made for: the learning processor is optionally coupled for learning purposes with an input unit via which one or more interference maps can be read. In this way, a typical interference pattern is transmitted to the machine self-learning program, in particular for learning purposes. The learning processor can be used at any time without connecting the input unit after a learning phase and after learning is finished.

In order to obtain information about the respective process operation by the operator during the process control, an advantageous development of the device according to the invention is used, in which the learning processor is coupled to an output unit, by means of which the identification of one of the disturbance sources or the assignment of the analyzed defect map to a disturbance map can be visualized. This output unit can here advantageously be coupled wirelessly to the learning processor and can be any type of instrument with a display thereon.

In order to obtain a diagnostic system that is as autonomous as possible, provision is additionally made for: the learning processor has a neural network for performing a machine self-learning procedure. This enables the analysis of large data volumes of yarn tension measurement signals by artificial intelligence.

In order to monitor a plurality of processing stations in a deformation machine, the device according to the invention can advantageously be used in a development in which the learning processor is spatially separated from the input unit and the output unit. In this case, there is a possibility that: the learning processor is connected to a plurality of input units and in particular to a plurality of output units. The connection can be designed to be wireless, so that the learning processor can also be formed in a virtual space, for example.

The device variant of the invention, in which the diagnostic unit is connected to a machine control unit, by means of which control commands for changing the process can be executed, is advantageously used for automation. This enables, for example, the winding of a thread knot at the beginning or end of a wound bobbin after a thread knot has been identified.

Drawings

The method according to the invention for monitoring a deformation process will be explained in detail below with reference to the drawing with the aid of some embodiments of the device according to the invention. In the drawings:

figure 1 schematically illustrates one embodiment of a texturing process for making a crimped yarn;

FIG. 2 schematically illustrates one embodiment of the apparatus of the present invention for monitoring the deformation process shown in FIG. 1;

FIG. 3 shows schematically a yarn tension diagram comprising a sequence of measurement signals;

FIG. 4 schematically illustrates a further embodiment of the apparatus of the present invention for monitoring the deformation process shown in FIG. 1;

fig. 5.1 schematically shows an analysis diagram of a time series with a plurality of yarn tension measurement signals;

fig. 5.2 schematically shows a defect map with a time series of a plurality of yarn tension measurement signals;

fig. 6.1 to 6.5 schematically show a plurality of interference patterns with a respective one of the measuring signal profiles of the yarn tension of different interference sources.

Detailed Description

Fig. 1 schematically shows an exemplary embodiment of a texturing process for producing a crimped yarn. In this case, the deformation process is described by a processing station of the deformation machine. Such texturing machines usually have a plurality of processing stations in order to perform a texturing process on a plurality of synthetic threads in parallel. Fig. 1 schematically illustrates a processing station 1 and a bobbin station 2 of such a texturing machine. The processing station 1 has a creel 4 in which a supply bobbin 5 and a reserve bobbin 6 are held. The supply bobbin 5 provides a yarn 3 which is conveyed in the processing station 1 for drawing and texturing. The yarn end of the supply bobbin 5 is interconnected with the yarn start of the reserve bobbin 6 by a yarn knot. This achieves a continuous withdrawal of the yarn 3 after the supply bobbin 5 has run out. The yarn end of the reserve bobbin 6 is then connected to the yarn start of a new supply bobbin 5.

The yarn is drawn off from the feed bobbin 5 by the first transport mechanism 7.1. The transport mechanism 7.1 is driven via a drive 8.1. The transport mechanism 7.1 is formed in this exemplary embodiment by a driven thread guide roller and a freely rotatable roller, which are multiply wound with the thread. In the further course of the thread, a heating device 9, a cooling device 10 and a texturing unit 11 are arranged downstream of the transport device 7.1. The deformation assembly 11 is driven by a deformation drive 11.1. The texturing unit 11 is preferably designed as a friction false twister in order to generate a false twist in the yarn, which false twist generates a crimping effect of the individual threads of the yarn.

For drawing the thread, a second transport mechanism 7.2 is provided downstream of the texturing unit 11, which is driven by a drive 8.2. The transport device 7.2 is identical in construction to the first transport device 7.1, the second transport device 7.2 being operated at a higher peripheral speed for drawing the thread. The synthetic thread 3 is thus deformed and simultaneously drawn in the processing station 1. After the yarn 3 has been crimped, it is guided by a third transport mechanism 7.3 to a bobbin station 2. The transport mechanism 7.3 is driven by a drive 8.3.

The bobbin station 2 has a bobbin holder 13, which holds a bobbin 14. The bobbin holder 13 is pivotably designed and can be operated manually or automatically for changing the bobbin 14. The bobbin holder 13 is provided with a drive roller 15 which is driven by a roller drive 15.1. In order to lay the yarn on the outer circumference of the bobbin 14, the bobbin station 2 is provided with a traversing unit 12 having a drivable yarn traversing guide. For this purpose, the yarn traverse guide is driven in an oscillating manner by a traverse drive 12.1.

The traversing drive 12.1 and the roller drive 15.1 of the bobbin station 2 are designed as independent drives and are connected to a machine control unit 16. The drives 8.1, 8.2 and 8.3 of the transport devices 7.1, 7.2 and 7.3 of the processing station 1 and the deformation drive 11.1 of the deformation aggregate 11 are likewise designed as independent drives and are coupled to the machine control unit 16.

In order to monitor the texturing process, the yarn tension is continuously measured on the moving yarn 3 in a measuring point between the transport devices 7.2 and 7.3. For this purpose, a thread tension measuring device 17 is provided, which has a thread tension sensor 17.1 and a measuring signal receiver 17.2. The thread tension measuring device 17 is connected to a data evaluation device 18 in the form of a diagnostic unit. For further explanation of the diagnostic unit 18, additional reference is made to fig. 2.

Fig. 2 schematically shows a diagnostic unit 18 for evaluating the yarn tension measurement signal. The diagnostic unit 18 in this embodiment includes a learning processor 20. The learning processor 20 is directly connected to the measurement signal receiver 17.2 of the yarn tension measuring device 17. The learning processor 20 is designed to be programmable and preferably has a neural network for executing a machine self-learning procedure. The machine self-learning program comprises at least one machine learning algorithm in order to be able to carry out an extensive analysis of the yarn tension measurement signal with artificial intelligence.

The learning processor 20 is provided with an input unit 22 and an output unit 23. The connection between the learning processor 20 and the yarn tension measuring device 17 and between the input unit 22 and the output unit 23 can be realized by a respective wired or wireless connection. In particular in the case of wireless connections, the following possibilities exist: it is not necessary to hold the units directly on the texturing machine. This also makes use of the self-learning procedure existing in the virtual space. This has the following possibilities: the learning processor 20 is provided independently of the input unit 22 and the output unit 23.

The yarn tension measurement signal transmitted by the measurement signal receiver 17.1 is analyzed in a learning processor 20 by means of a machine self-learning program. The machine self-learning program has at least one machine learning algorithm which, for the early diagnosis of one of the interference sources, performs a structured evaluation of the sequence of yarn tension measurement signals occurring in a time interval by means of a neural network. At the same time, the chronological changes in the measurement signals of the yarn tension measurement signals are evaluated in order to reveal typical features for identifying a specific source of disturbance.

In the embodiment shown in fig. 2, the input unit 22 is configured to the learning processor 20. In this regard, the machine learning algorithm can continuously learn using the interference patterns of known interference sources. Usually, after a run time, the learning phase of the machine learning algorithm reaches a state of maturity and thus an end-of-learning operating state. In this state the diagnostic unit 18 is operating without the input unit 22. Accordingly, the diagnostic unit 18 can also be used in known deformation processes which have a self-learning program with a learned machine algorithm from the beginning. The use of the input unit then becomes unnecessary.

The self-learning program of the learning processor can be replaced or relearned at a central location for the presence of unknown sources of interference in a learned system, which are indistinguishable by machine learning algorithms. The flaw map of unknown sources of interference is first directly transmitted to a central learning location so that existing machine learning algorithms can already be made to learn.

Fig. 3 shows a time-series example of a yarn tension measurement signal in a graph. The height T of the yarn tension is plotted on the ordinate of the graph, and the time T is plotted on the abscissa. The profile of the measurement signal enables a sudden increase in the yarn tension to be detected only from time t₁To t₂Occurs within a short time. The yarn tension measurement signals before and after the increase are not noticeable. This measuring signal sequence indicates a source of interference in the form of a thread knot which has passed the measuring point. In order to identify such interference sources, a typical measurement signal profile is transmitted to a self-learning program in advance, so that such characteristic features can be recognized during a later evaluation of the measurement signal sequence.

In the case of a monitoring of the course of the texturing process in which a thread knot in the thread passes through the measuring point, a similar measuring signal sequence is generated by the measuring signal receiver 17.1 and transmitted to the diagnostic unit 18. By analyzing the measurement signal sequence in a learning processor, typical, representative measurement signal variations are identified and relevant interference sources are identified. In this case, it is not important whether the yarn tension variation exceeds a threshold value or remains within an allowable tolerance range.

Since not only product defects appear as interference sources in a deformation process, a differentiated and in particular expanded analysis and diagnosis of interference sources is desirable. For this purpose, fig. 4 shows a further exemplary embodiment of the device according to the invention, as it can be used, for example, in the deformation process shown in fig. 1. In this case, the diagnostic unit 18 comprises a yarn tension evaluation unit 19 which is directly connected to the yarn tension measuring device 17. The measurement signal of the measurement signal receiver 17.2 is thus transmitted to the yarn tension evaluation unit 19. The time course of the first measurement signal within the yarn tension evaluation unit 19 is recorded in sequence and generated as an analysis chart. At the same time, the measurement signal is compared with a threshold value. The yarn tension measurement signal is usually compared with an upper and a lower limit value for the yarn tension. When an inadmissible tolerance deviation of the yarn tension is detected, a short-term measurement signal curve of the yarn tension is recorded and generated as a defect map.

Depending on whether there is an analysis map without a threshold crossing or a defect map with a threshold violation, the map is passed to a learning processor 20. The learning processor 20 is adapted accordingly with respect to its machine learning algorithm in order to be able to carry out a corresponding analysis for diagnosing the cause of the disturbance.

Fig. 5.1 and 5.2 show an analysis diagram without threshold violation and a defect diagram with threshold violation, respectively. The yarn tension measurement signal is compared with an upper limit value and a lower limit value. In the analysis and defect maps the upper limit values are marked with the marking letter OG and the lower limit values are marked with the marking letter UG. For this purpose, the yarn tension T is plotted on the ordinate and the time T is plotted on the abscissa. In the signal curve of the yarn tension measurement signal shown in fig. 5.1, the exceeding of the upper limit value and the falling below the lower limit value cannot be detected. In this regard, the signal variation of the yarn tension caused by a source of disturbance is within the permitted range. Nevertheless, for example, a thread knot may also form an impermissible defect in the fabric in a subsequent process.

In the defect map shown in fig. 5.2, the yarn tension measurement signal exceeds the upper limit value OG for a short time, so that there is a threshold violation and therefore an impermissible deterioration of the yarn quality. The inventive method and the inventive device are therefore independent of whether process disturbances caused by a disturbance source lead to permissible or impermissible changes in the thread tension.

As can be seen from the diagram of fig. 4, a plurality of so-called interference maps are specified in advance for the learning processor via the input unit. A disturbance map contains a yarn tension measurement signal curve that is representative for a certain disturbance source. This allows the machine self-learning procedure to learn and expand to make it possible to unambiguously diagnose and identify sources of interference.

Fig. 6.1 to 6.5 show, by way of example, a plurality of interference patterns of different interference sources schematically by way of a typical yarn tension measuring signal sequence. Each of the interference maps represents a typical source of interference, as it may occur during the warping process. The interference diagram according to fig. 6.1 shows the signal curve of the yarn tension in the case of a yarn knot, which is designated here as interference source a.

The source of disturbance B shown in fig. 6.2 shows a sudden drop in the yarn tension, as may be caused, for example, by a yarn break or an operator error. In this case, the yarn tension on the yarn tension sensor suddenly drops completely sharply.

The interference graph shown in fig. 6.3 determines the interference source C. The disturbance source C may represent, for example, wear on one of the processing assemblies or on a thread guide element.

In the interference pattern shown in fig. 6.4, a brief loss of yarn tension occurs, which is assigned to the interference source D. In this case, there may be a process disturbance in the bobbin station during the bobbin change.

The repeated increase in the measurement signal of the yarn tension can be seen in the interference diagram shown in fig. 6.5. The corresponding interference pattern belongs in this case to the interference source E. The source of disturbance E may for example be an uneven deformation of the yarn.

In this regard, any of the interference sources a to E corresponds to a defined process interference or a defined malfunction or a defined product defect, respectively. The interference diagrams shown in fig. 6.1 to 6.5 and their sources of interference are exemplary. In a modification process as shown in fig. 1, a number of interference sources may occur, which are based on a specific measurement signal sequence for diagnostic discrimination.

In the exemplary embodiment of the device according to the invention shown in fig. 4, the diagnostic unit 18 is connected directly to a machine control unit 16 via a learning processor 20. As shown in fig. 1, the machine control unit 16 is coupled to drives and actuators of the processing assemblies of the processing station 1, which are controlled and adjusted to carry out the deformation process. In this regard, process intervention can be performed directly after identifying and diagnosing an interference source. This makes it possible, for example, to initiate a bobbin change in the bobbin station 2 when the source of interference a due to a yarn knot is diagnosed, so that the yarn knot is wound on the end or start of a bobbin. Also, the corresponding maintenance work can be started by diagnosing the wear phenomenon in advance. This allows the processing unit to be repaired and maintained without rain, in order to keep the quality as stable as possible during the production of the crimped thread. The embodiment of the apparatus of the invention shown in fig. 4 is still in the learning phase. Immediately after the learning of the machine learning algorithm to identify all known sources of interference has ended, the diagnostic unit 18 can be run without the input unit 22. The yarn tension signal is then transmitted directly to the learning processor 20 and directly analyzed.