DEVICE AND METHOD FOR DETECTING A FAULT IN A SPINNING MILL AND FOR ESTIMATING ONE OR MORE SOURCES OF THE FAULT

Abstract

An electronic device and associated method are used to detect a fault in a spinning mill and to estimate one or more sources of the fault, the spinning mill including a plurality of textile machines that sequentially process textile materials. With the electronic device, the method receives parameter information of one or more of the textile machines and of one or more of the textile materials. The electronic device detects faults and location of the faults by identifying parameter information of the textile materials deviating from reference information. The electronic device is used to access configuration information of the textile machines and knowledge-based information related to possible sources of faults in the spinning mill. The method incudes using the electronic device to apply parameter information, configuration information, and knowledge-based information to one or more machine-learning algorithms to estimate the sources of the faults.

Claims

1-20. (canceled)

21. An electronic device for detecting a fault in a spinning mill and for estimating one or more sources of the fault, wherein the spinning mill includes a plurality of textile machines that sequentially process textile materials, wherein the electronic device is configured to perform the following: receive parameter information of one or more of the textile machines and of one or more of the textile materials; detect faults and location of the faults by identifying parameter information of the textile materials deviating from reference information; access configuration information of the textile machines; access knowledge-based information related to possible sources of faults in the spinning mill; and apply parameter information, configuration information, and knowledge-based information to one or more machine-learning algorithms to estimate the sources of the faults.

22. The electronic device according to claim 21, wherein to identify the parameter information of the textile materials deviating from the reference information, the electronic device is further configured to compare a difference between one or more parameter values and one or more reference values to a threshold.

23. The electronic device according to claim 21, further configured to use the configuration information to determine one or more possible or likely sequences of the textile machines between the location of the fault and the sources of the fault.

24. The electronic device according to claim 21, further configured to use the configuration information to determine one or more impossible or unlikely sequences of the textile machines between the location of the fault and the sources of the fault.

25. The electronic device according to claim 21, further configured to use one or more of time-dependent parameter information and time-dependent configuration information.

26. The electronic device according to claim 21, further configured to apply one or more machine-learning algorithms selected from a Linear Regression technique, a Logistic Regression technique, a Support Vector Machine, a Decision Tree, a Random Forest technique, a K-Nearest Neighbors Algorithm, a K-Means Clustering technique, a Naïve Bayes classifier, and a Principal Component Analysis technique, while taking into account that one or more of the parameter information, the configuration information, and the knowledge-based information may be incomplete or incorrect.

27. The electronic device according to claim 26, furthered configured to request or access supplemental configuration information, and apply the supplemental configuration information to the one or more machine-learning algorithms.

28. The electronic device according to claim 21, further configured to determine if the textile machines or components of the textile machines related to the sources of the fault must be replaced, repaired, modified, or adjusted differently.

29. The electronic device according to claim 28, further configured to determine if maintenance work on the textile machines or the components of the textile machines related to the sources of the fault will result in a downtime of the textile machine or the spinning mill.

30. The electronic device according to claim 28, further configured to receive feedback information indicating if maintenance work on the textile machines or the components of the textile machines related to the sources of the fault resulted in correction of the fault, and to update the knowledge-based information accordingly.

31. A method for detecting a fault in a spinning mill and for estimating one or more sources of the fault, the spinning mill including a plurality of textile machines that sequentially process textile materials, the method comprising: with an electronic device, receiving parameter information of one or more of the textile machines and of one or more of the textile materials; with the electronic device, detecting faults and location of the faults by identifying parameter information of the textile materials deviating from reference information; with the electronic device, accessing configuration information of the textile machines; with the electronic device, accessing knowledge-based information related to possible sources of faults in the spinning mid; and with the electronic device, applying parameter information, configuration information, and knowledge-based information to one or more machine-learning algorithms to estimate the sources of the faults.

32. The method according to claim 31, comprising identifying the parameter information of textile materials deviating from reference information by comparing with a threshold a difference between one or more parameter values and one or more reference values.

33. The method according to claim 31, further comprising using the configuration information to determine one or more possible or likely sequences of the textile machines between the location of the fault and the sources of the fault.

34. The method according to claim 31, further comprising using the configuration information to determine one or more impossible or unlikely sequences of the textile machines between the location of the fault and the sources of the fault.

35. The method according to claim 31, further comprising using one or more of time-dependent parameter information and time-dependent configuration information.

36. The method according to claim 31, further comprising applying one or more machine-learning algorithms selected from a Linear Regression technique, a Logistic Regression technique, a Support Vector Machine, a Decision Tree, a Random Forest technique, a K-Nearest Neighbors Algorithm, a K-Means Clustering technique, a Naïve Bayes classifier, and a Principal Component Analysis technique, while taking into account that one or more of the parameter information, the configuration information, and the knowledge-based information may be incomplete or incorrect.

37. The method according to claim 31, further comprising requesting or accessing supplemental configuration information, and applying the supplemental configuration information to the one or more machine-learning algorithms.

38. The method according to claim 31, further comprising determining if the textile machines or components of the textile machines related to the sources of the fault must be replaced, repaired, modified, or adjusted differently.

39. The method according to claim 38, further comprising determining if maintenance work on the textile machines or the components of the textile machines related to the sources of the fault will result in a downtime of the textile machine or the spinning mill.

40. The method according to claim 38, further comprising receiving feedback information indicating if maintenance work on the textile machines or the components of the textile machines related to the sources of the fault resulted in correction of the fault, and updating the knowledge-based information accordingly.

Description

BRIEF DESCRIPTION OF DRAWINGS

[0045] The invention will be better understood with the aid of the description of an embodiment given by way of example an illustrated by the figures, in which:

[0046] FIG. 1 illustrates schematically exemplary textile machines of a spinning mill for processing input textile materials into output textile materials;

[0047] FIG. 2 illustrates schematically an exemplary configuration of a spinning mill, and a device for detecting a fault in the spinning mill and for estimating one or more sources of the fault;

[0048] FIG. 3 illustrates schematically an exemplary configuration of a spinning mill, a detected fault, two possible sources of the fault, and a device for detecting the fault in the spinning mill and for estimating the two possible sources of the fault;

[0049] and

[0050] FIG. 4 illustrates schematically possible method steps of a method for detecting that a spinning mill has a fault and for estimating one or more sources of the fault.

DETAILED DESCRIPTION OF THE INVENTION

[0051] Reference will now be made to embodiments of the invention, one or more examples of which are shown in the drawings. Each embodiment is provided by way of explanation of the invention, and not as a limitation of the invention. For example features illustrated or described as part of one embodiment can be combined with another embodiment to yield still another embodiment. It is intended that the present invention include these and other modifications and variations to the embodiments described herein.

[0052] FIG. 1 illustrates schematically exemplary textile machines 12, 23, 34, 45, 56, 67, 78 of a spinning mill for processing input textile materials 1, 2, 3, 4, 5, 6, 7 into output textile materials 2, 3, 4, 5, 6, 7, 8. Depending on the spinning mill, different types of textile machines 12, 23, 34, 45, 56, 67, 78 and/or different sequences of textile machines 12, 23, 34, 45, 56, 67, 78 may be involved. In the example illustrated in FIG. 1, one or more blow room textile machines 12 are arranged for processing raw cotton 1 into chute mat 2. One or more carding textile machines 23 are arranged for processing chute mat 2 into carded sliver 3. One or more breaker draw frame textile machines 34 are arranged for processing carded sliver 3 into break drawn sliver 4. One or more finisher draw frame textile machines 45 are arranged for processing break drawn sliver 4 into finisher draw sliver 5. One or more speed frame textile machines 56 are arranged for processing finisher draw sliver 5 into roving 6. One or more ring frame textile machines 67 are arranged for processing roving 6 into ring cops 7. One or more winding textile machines 78 are arranged for processing ring cops 7 into yarn cones 8.

[0053] A spinning mill enables producing from a source textile material 1 a desired textile material 8 in a desired quantity and/or quality. Each of the textile machines 12, 23, 34, 45, 56, 67, 78 illustrated in FIG. 1 is configured for processing in accordance to a desired quantity and/or quality input textile materials 1, 2, 3, 4, 5, 6, 7 into output textile materials 2, 3, 4, 5, 6, 7, 8. Quantity and/or quality of output textile materials 2, 3, 4, 5, 6, 7, 8 depend on parameters of the textile machines 12, 23, 34, 45, 56, 67, 78 and/or on parameters of the input textile materials 1, 2, 3, 4, 5, 6, 7. Quantity and/or quality of output textile materials 2, 3, 4, 5, 6, 7, 8 may not only depend on parameters of the direct input textile material 1, 2, 3, 4, 5, 6, 7 of a particular textile machines 12, 23, 34, 45, 56, 67, 78, but may also depend on parameters of any earlier textile material 2, 3, 4, 5, 6 produced by any earlier textile machine 12, 23, 34, 45, 56, as well as on parameters of the source textile material 1. For example, the direct input textile material 4 of the finisher draw frame 45 is break drawn sliver 4, while the earlier textile materials also include the raw cotton 1, the chute mat 2, and the carded sliver 3. Thus, quantity and/or quality of an output textile material 2, 3, 4, 5, 6, 7, 8 may not only be degraded if quantity and/or quality of the direct input textile material 1, 2, 3, 4, 5, 6, 7 does not match a desired quantity and/or quality, quantity and/or quality of an output textile material 3, 4, 5, 6, 7, 8 may also be degraded if quantity and/or quality of an earlier input textile material 1, 2, 3, 4, 5, 6 does not match a desired quantity and/or quality. For example, quantity and/or quality of the finisher draw sliver 5 produced by the draw frame textile machines 45 may be degraded, if quantity and/or quality of one or more of the raw cotton 1, the chute mat 2, and the carded sliver 3 does not match a desired quantity and/or quality.

[0054] FIG. 2 illustrates schematically an exemplary configuration of a spinning mill M. Furthermore, FIG. 2 illustrates schematically a device E for detecting a fault in the spinning mill and for estimating one or more sources eS1, eS2 of the fault. The spinning mill M illustrated in FIG. 2 includes different and/or additional textile machines than the textile machines illustrated in FIG. 1. The present disclosure is not limited to the exemplary configuration of a spinning mill M illustrated in FIG. 2, but also applies to any other configuration of a spinning mill.

[0055] The spinning mill M illustrated in FIG. 2 includes a bale opener 12 for processing raw material 1 into so-called microtufts 2. The bale opener 12 is followed by a pre-cleaner 23 for processing microtufts 2 into pre-cleaned textile material 3. The pre-cleaner 23 is followed by a homogenous mixer 34 for processing pre-cleaned textile material 3 into mixed textile material 4. The homogenous mixer 34 is followed by a storage and feeding machine 45 for finely cleaning the mixed textile material 4 into finely cleaned textile material 5. The storage and feeding machine 45 is followed by a condenser 56 for additionally cleaning the finely cleaned textile material 5 into additionally cleaned textile material 6. The condenser 56 is followed by four draw frame and sliver coiler textile machines 67.1, 67.2, 67.3, 67.4 for processing the additionally cleaned textile material 6 into card sliver coils 7. The four draw and sliver coiler units 67.1, 67.2, 67.3, 67.4 are followed by eight speed frame textile machines 78.1, 78.2, 78.3, 78.4, 78.5, 78.6, 78.7, 78.8 for processing the card sliver cols 7 into the roving 8. The eight speed frame machines 78.1, 78.2, 78.3, 78.4, 78.5, 78.6, 78.7, 78.8 are followed by two ring frame textile machines 89.1, 89.2 for processing the roving 8 into ring cops 9.

[0056] As illustrated in FIG. 2, transport systems are arranged for transporting textile materials from a previous textile machine to a next textile machine. The transport systems may include duct systems D, trolley systems T, rail systems R, etc. For example, in a duct system D textile material is transported by means of an air stream generated by a ventilator. For example, in a trolley system T textile material is transported by means of containers arranged on trolleys. For example, in a rail system R textile material is transported by means of transport devices arranged on rails. Other transport systems may be involved. In the example illustrated in FIG. 2, a duct system D is arranged for transporting textile material through the sequence of textile machines which include the bale opener 12, the pre-cleaner 23, the homogenous mixer 34, the storage and feeding machine 45, the condenser 56, and the draw frame and sliver coiler textile machines 67.1, 67.2, 67.3, 67.4. A trolley system T is arranged for transporting textile material from the draw and sliver coiler units 67.1, 67.2, 67.3, 67.4 to the speed frame textile machines 78.1, 78.2, 78.3, 78.4, 78.5, 78.6, 78.7, 78.8. A rail system R is arranged for transporting textile material from the speed frame textile machines 78.1, 78.2, 78.3, 78.4, 78.5, 78.6, 78.7, 78.8 to the ring frame textile machines 89.1, 89.2. Transport systems may be arranged in a different manner.

[0057] As illustrated in FIG. 2, parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9 is captured. The parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9 may include parameter values captured from the textile machines 12, 23, 34, 45, 56, 67, 78, 89 and/or parameter values captured from the textile materials 1, 2, 3, 4, 5, 6, 7, 8, 9. FIG. 2 schematically illustrates exemplary locations for capturing parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9. However, in particular because of complexity and/or costs, in a typical configuration of a spinning mill M, one or more of the parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9 of the textile machines 12, 23, 34, 45, 56, 67, 78, 89 and/or textile materials 1, 2, 3, 4, 5, 6, 7, 8, 9 may only be partially captured or not at all captured.

[0058] The parameter values of the parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9 may relate to a rotation speed, a power consumption, a maintenance date, etc. of respective textile machines 12, 23, 34, 45, 56, 67, 78, 89. The parameter values of the parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9 may relate to a thickness, a weight, etc. of respective textile materials 1, 2, 3, 4, 5, 6, 7, 8.

[0059] For capturing parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9, respective electronic sensors are arranged for capturing respective parameter values.

[0060] Electronic sensor may relate to electronic sensors for capturing a rotation speed, electronic sensors for capturing power consumption, etc. For capturing parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9, laboratory-analyzed findings may be required for capturing respective parameter values. Laboratory-analyzed findings may relate to fiber densities, lubricant quality, etc.

[0061] Parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9 may be captured regularly or irregularly in time. Parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9 may be captured within short intervals at high speed or within long intervals at low speed. For example, parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9 captured with electronic sensors may be captured regularly at high speed such as every minute, every second, etc. For example, parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9 requiring laboratory-analyzed findings may be captured irregularly at low speed, such as every Monday and Thursday, after machine maintenance, etc.

[0062] As illustrated in FIG. 2, the parameter information p1, p2, p3, p4, p5, p7, p8, p9 is captured at specific locations. Accordingly, specific parameter information, such as parameter information with reference sign p3, relates to a specific location in the spinning mill, in particular a specific rotation speed of a rotor of a textile machine, a specific power consumption of a specific textile machine, a specific diameter of a textile material produced by a specific textile machine, etc.

[0063] FIG. 2 schematically illustrates an electronic device E for detecting a fault in a spinning mill and for estimating one or more sources eS1, eS2 of the fault in accordance to this disclosure.

[0064] The electronic device E may take the form of a computer, for example a computer that is generally used in one place (such as a conventional desktop computer, workstation, server, etc.) as well as a computer that is generally portable (such as a laptop, notebook, tablet, handheld computer, etc.). The electronic device E may include a machine-readable medium having stored thereon instructions which program a processor of the electronic device E to perform some or all of the operations and functions described in this disclosure. A machine-readable medium may include any mechanism for storing or transmitting information in a form readable by a machine (e.g., a computer), such as Hard Disk drives (HD), Solid State Disk drives (SSD), Compact Disc Read-Only Memory (CD-ROMs), Read-Only Memory (ROMs), Random Access Memory (RAM), Erasable Programmable Read-Only Memory (EPROM), etc. In other embodiments, some of these operations and functions might be performed by specific hardware components that contain hardwired logic. Those operations and functions might alternatively be performed by any combination of programmable computer components and fixed hardware circuit components. In one embodiment, the machine-readable medium includes instructions stored thereon, which when executed by a processor, causes the processor to perform the method on an electronic device E as described in this disclosure.

[0065] As illustrated in FIG. 2, the electronic device E is configured to receive parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9 of the textile machines 12, 23, 34, 45, 56, 67, 78, 89 and/or of the textile materials 1, 2, 3, 4, 5, 6, 7, 8, 9. For example, electronic sensors configured to capture respective parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9 are arranged at respective locations in the spinning mill M and electronic signals representing the captured parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9 are transmitted from the electronic sensor via a computer network to the electronic device E. Furthermore, parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9 captured in laboratory-analyzed findings may be transmitted via a computer network to the electronic device E. Accordingly, the electronic device E is configured to receive parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9 which includes a wide variety of parameter types captured at a wide variety of locations.

[0066] The electronic device E is configured to detect the fault in the spinning mill by identifying parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9 of textile materials 1, 2, 3, 4, 5, 6, 7, 8, 9 deviating from reference information. For example, the difference between one or more parameter values and one or more reference information is compared with a threshold, and the fault in the spinning mill M is detected if the difference exceeds the threshold. For example, the fault in the spinning mill M is detected if the diameter of a textile material exceeds a maximal diameter, or if the diameter of the textile material is below a minimal diameter. Together with detecting the fault, the location of the fault is detected, because each of the parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9 is captured at a specific location in the spinning mill M.

[0067] As illustrated in FIG. 2, the electronic device E is configured to access configuration information cI related to a configuration of the spinning mill M. As illustrated in FIG. 2, the configuration information cI may be stored externally from the electronic device E, such as in a database of a computer server having stored information about the spinning mill M. Alternatively or additionally, the configuration information cI may be stored in a machine-readable medium of the electronic device E. The configuration information cI may include information about types of textile machines, model information about textile machines, locations of textile machines, ordering of textile machines, operation times of textile machines, etc.

[0068] The configuration information cI enables determining or estimating for any textile material at any location in the spinning mill M the sequence of textile machines which was involved for producing the textile machine. For example, the sequence of textile machines involved for producing a specific cop 9 may include the ring frame textile machine with reference numeral 89.1 which produced the specific cop 9, the speed frame textile machine with reference numeral 78.3, for example because the configuration information cI has stored information that during the relevant time period, textile material produced by the speed frame textile machine with reference numeral 78.3 was transported to the ring frame textile machine with reference numeral 89.1. Furthermore, in this example the configuration information cI indicates that the sequence of textile machines for producing the specific cop 9 includes the draw and sliver coiler unit with reference numeral 67.4, for example because the configuration information cI has stored information that during the relevant time period, textile material produced by draw and sliver coiler unit with reference numeral 67.4 was transported to the speed frame textile machine with reference numeral 78.3. Furthermore, in accordance to the configuration illustrated in FIG. 2, the sequence of textile machines involved for producing the specific cop 9 includes the condenser 56, the storage and feeding machine 45, the homogenous mixer 34, the pre-cleaner 23, and the bale opener 12, for example because the spinning mill M does not include any alternative sequence of textile machines as illustrated in FIG. 2. Thus, if a fault is detected in the specific cop 9 produced by the ring frame machine with reference numeral 89.1, the one or more sources eS1, eS2 of the fault are limited to the ring frame machine with reference numeral 89.1, the speed frame textile machine with reference numeral 78.3, the draw and sliver coiler unit with reference numeral 76.4, the condenser 56, the storage and feeding machine 45, the homogenous mixer 34, the pre-cleaner 23, and the bale opener 12, while the ring frame machine with reference numeral 89.2, the speed frame textile machines with reference numerals 78.1, 78.2, 78.4, 78.5, 78.6, 78.7, 78.8, the draw and sliver coiler units with reference numerals 67.1, 67.2, 67.3 are excluded as possible sources of the fault.

[0069] Furthermore, the configuration information cI may include time dependent information, for example in accordance to a production plan involving different textile machines such as during the night and during the day, or such as during maintenance work and during normal production. Thus, for a first point in time it may be necessary to include respectively exclude a different set of textile machines as possible sources of the fault than for a second point in time. The configuration information cI may relate to actual values as well as past values. The configuration information cI may include information about machine settings, such as mechanical settings, technological settings, software settings, etc.

[0070] The configuration information cI may include indications if stored information may be incomplete or incorrect. For example, as schematically illustrated in FIG. 2, a trolley system T may be arranged for transporting textile material from frame and sliver coiler textile machines 67.1, 67.2, 67.3, 67.4 to speed frame textile machines 78.1, 78.2, 78.3, 78.4, 78.5, 78.6, 78.7, 78.8. The trolley system may require manual operation by operating personal, wherein operation personnel is instructed to manually record source and destination. Operation personnel may forget to record source and destination, or may record an incorrect source or destination. Accordingly, configuration information cI may include incomplete or incorrect information.

[0071] As illustrated in FIG. 2, the electronic device E is configured to access knowledge-based information kI related to knowledge about sources of faults in the spinning mill M. Knowledge about sources of faults in the spinning mill M may relate to long term information gathered by experts in the field of spinning mills. For example, knowledge information may include that in case cops 9 include a yarn having a thickness which varies more than a predefined reference variance the source of the fault is more likely located in a frame and sliver coiler textile machine 67.1, 67.2, 67.3, 67.4 than in a bale opener 12 or storage feeding machine 45, for example. Knowledge information may include information gathered from computer algorithms and artificial intelligence of a specific spinning mill or of a plurality of spinning mills. The plurality of spinning mills may include spinning mills of essentially the same configurations, of spinning mills of similar configurations, of spinning mills of different configurations, etc.

[0072] The electronic device E is configured to apply parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9, configuration information cI, and knowledge-based information kI to one or more machine-learning algorithms for estimating the one or more sources of the fault. Thus, information which includes parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9, configuration information cI, and knowledge-based information k is processed using one or more machine-learning algorithms.

[0073] The machine-learning algorithms may include one or more of a Linear Regression technique, a Logistic Regression technique, a Support Vector Machine, a Decision Tree, a Random Forest technique, a K-Nearest Neighbors Algorithm, a K-Means Clustering technique, a Naïve Bayes classifier, and a Principal Component Analysis technique.

[0074] A Linear Regression technique may detect dependencies within the information. If a fault is detected, or in other words if one of the parameter values deviates from a reference value, a Linear Regression technique may determine other parameter values which still do not deviate from reference values, but have a correlation with the parameter value which deviated from the reference value, thereby enabling estimating the one or more sources of the fault.

[0075] Similar to a Linear Regression technique, a Logistic Regression technique may detect dependencies within information related to binary values.

[0076] A Support Vector Machine may classify complex information by defining a hyperplane within the space defined by the information. For example, if currently analyzed information is on one side of the hyperplane, it may be estimated that a specific source of the fault is more likely than in case the information is on the other side of the hyperplane.

[0077] A Decision Tree may be applied to the information in order to estimate the one or more sources of the fault. In accordance to a Decision Tree, information is stepwise analyzed in order to estimate the one or more sources of the fault.

[0078] A Random Forest technique may be applied to the information in order to estimate the one or more sources of the fault. In accordance to a Random Forest technique, the information is applied to a collection of decision trees in order to estimate the one or more sources of the fault. A random forest is based on splitting features or information into a subset of trees. The model considers only a small subset of features or information rather than all of the features or information. A random forest can be executed in parallel on different computing engines, such as different cores of a processor, different processors of a computer system, different computer systems, etc., thereby enabling estimating the one or more sources of the fault within a predefined time period, in particular in case of high complexity. A random forest can handle different classes of features or information, such as binary features or information, categorical features or information, numerical features or information, etc. A random forest enables estimating the one or more sources of the fault without or with little pre-processing of the features or information and there is no need to rescale or transform the features or information. A random forest is particularly well suited in case of high dimensional data, features or information because execution is limited to subsets of data, features or information.

[0079] A K-Nearest Neighbors Algorithm may be applied to the information in order to estimate the one or more sources of the fault. In accordance to a K-Nearest Neighbors Algorithm, the information is applied to classes which have been defined on the basis of sample information and K nearest neighbors of the samples.

[0080] A K-Means Clustering technique may be applied to the information in order to estimate the one or more sources of the fault. In accordance to a K-Means Clustering technique, the information is applied to clusters which have been defined on the basis of sample information and an optimization of centroid position.

[0081] A Naïve Bayes classifier may be applied to the information in order to estimate the one or more sources of the fault. In accordance to a Naïve Bayes classifier, the information is classified in accordance to the Bayes rule.

[0082] A Principal Component Analysis technique may be applied to the information in order to estimate the one or more sources of the fault. In accordance to a Principal Component Analysis technique, a set of correlated variables is converted to a set of uncorrelated variables in order to remove redundancies.

[0083] Machine-learning algorithms may include generic algorithms such as Artificial Neural Networks (best for data patterns), Convolutional Neural Networks (best for images), Recurrent Neural Networks (best for audio signals), Self-Organizing Maps (best for feature detection), Deep Boltzmann Machines (best for system modelling), AutoEncoders (best for property detection), etc.

[0084] One or more machine-learning algorithm may be applied to the information depending on the class or type of a spinning mill.

[0085] An Artificial Neural Network (ANN) may be applied to the information in order to estimate the one or more sources of the fault by executing regression and classification where patterns and sequences are recognized. In accordance to an Artificial Neural Network, information is stepwise analyzed in order to estimate the one or more sources of the fault.

[0086] A Convolutional Neural Network (CNN) may be applied to still image, video stream, visual and other two-dimensional data in order to estimate the one or more sources of the fault. In accordance to a convolutional neural network a mathematical operation called convolution is applied in order to find specific features to estimate the one or more sources of the fault.

[0087] A Recurrent Neural Network (RNN) may be applied to time series data, sequence modeling or audio signals (including noise patterns) in order to estimate the one or more sources of the fault. In accordance to Recurrent Neural Network, long short-term memory (LSTM) is used to process data within their internal state (memory) to process variable length sequences of inputs in order to estimate the one or more sources of the fault.

[0088] Self-Organizing Maps (SOM) may be applied to the information for reduction of the dimensionality and visual representation of data. In accordance with Self-Organizing Maps competitive learning is applied to approximate the data distribution in order to estimate the one or more sources of the fault.

[0089] A K-Means Clustering technique may be applied to cluster data into separate partitions in order to estimate the one or more sources of the fault. In accordance to a K-Means Clustering techniques are used for feature learning where the information is applied to clusters which have been defined on the basis of sample information and an optimization of centroid position

[0090] Deep Bolzmann Machines (DBM) may be applied to model and monitor the behavior of a spinning mill or its subsystem including climatic conditions in order to estimate the one or more sources of the fault. In accordance to a Deep Bolzmann Machine internal representations are learned to represent and solve difficult combinatoric problems in order to estimate the one or more sources of the fault.

[0091] AutoEncoders may be applied to information for dimensionality reduction (decoding) and information retrieval (encoding) in order to estimate the one or more sources of the fault. In accordance with AutoEncoders different input representations are learned to assume useful properties in order to estimate the one or more sources of the fault.

[0092] Artificial Neural Networks, Convolutional Neural Networks and Recurrent Neural Networks belong to the class of supervised machine-learning algorithms. Self-Organizing Maps, Deep Boltzmann Machines and AutoEncoders belong to the class of unsupervised machine-learning algorithms. Artificial Neural Networks may be used for regression and classification. Convolutional Neural Networks may be used for computer vision. Recurrent Neural Networks may be used for time series analysis. Self-Organizing Maps may be used for feature detection. Deep Boltzmann Machines and AutoEncoders may be used for recommendation systems.

[0093] The source of the fault may relate to wrongly adjusted textile machines 12, 23, 34, 45, 56, 67, 78, 89, to wear of components of textile machines 12, 23, 34, 45, 56, 67, 78, 89, etc.

[0094] Estimating the source of the fault may include a step of providing information if textile machines 12, 23, 34, 45, 56, 67, 78, 89 and/or components of textile machines 12, 23, 34, 45, 56, 67, 78, 89 must be replaced, repaired, modified, adjusted differently, etc.

[0095] FIG. 3 illustrates schematically an exemplary configuration of a spinning mill, a detected fault, two possible sources of the fault, and a device for detecting the fault in the spinning mill and for estimating the two possible sources of the fault.

[0096] The electronic device E receives parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9) related to one or more parameters of the textile machines 12, 23, 34, 45, 56, 67, 78, 89 and of one or more textile materials 1, 2, 3, 4, 5, 6, 7, 8, 9. For example, upon receipt of a new parameter value, the parameter value is compared to a reference value. For example, a diameter of the textile material is compared with a reference diameter. A fault is detected if the parameter deviates from the reference information. Because of information about the location or origin of the parameter value, such as the location of a sensor, the location of the fault is also detected. Thus, the electronic device E detects the fault dF in the spinning mill M and the location of the fault by identifying parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9 related to one or more parameters of one or more textile materials 1, 2, 3, 4, 5, 6, 7, 8, 9 which deviate from reference information. In the exemplary configuration illustrated in FIG. 3, the fault dF relates to cops 9 produced by the ring frame textile machine with reference numeral 89.1.

[0097] The electronic device E accesses configuration information cI related to the configuration of the spinning mill M and knowledge-based information kI related to knowledge about sources of faults in the spinning mill M and applies parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9, configuration information cI, and knowledge-based information kI to one or more machine-learning algorithms for estimating the one or more sources of the fault eS1, eS2. In the example illustrated in FIG. 3, two sources of the fault eS1, eS2 are estimated, wherein the first source of the fault with reference sign eS1 concerns the frame and sliver coiler textile machine with reference numeral 67.1, and the second source of the fault with reference sign eS2 concerns the speed frame textile machine with reference numeral 78.4. For example, a Random Forest technique was applied to parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9, configuration information cI and knowledge information kI and resulted in the estimation of the first source eS1 and the second source eS2.

[0098] FIG. 4 illustrates schematically possible method steps of a method for detecting that a spinning mill has a fault dF and for estimating one or more sources of the fault eS1, eS2. In step S1, an electronic device E receives parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9 of one or more textile machines (12, 23, 34, 45, 56, 67, 78, 89) and of one or more textile materials 1, 2, 3, 4, 5, 6, 7, 8, 9. In step S2, the electronic device E detects the fault dF and its location by identifying parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9 of textile materials 1, 2, 3, 4, 5, 6, 7, 8, 9 deviating from reference information. In step S3, the electronic device E accesses configuration information cI of textile machines 12, 23, 34, 45, 56, 67, 78, 89 of the spinning mill M. In step S4, the electronic device accesses knowledge-based information kI related to knowledge about possible sources eS' of possible faults dF′ in the spinning mill M. In step S5, the electronic device applies parameter information p1, p2, p3, p4, p5, p6, p7, p8, p9, con-figuration information cI, and knowledge-based information kI to one or more machine-learning algorithms for estimating the one or more sources eS1, eS2 of the fault dF.

REFERENCE NUMERALS/SIGNS

[0099] M spinning mill [0100] 1,2,3,4,5,6,7,8,9 textile materials [0101] 12,23,34,45,56,67,78,89 textile machines [0102] p1,p2,p3,p4,p5,p6,p7,p8,p9 parameters of the textile machines and/or the textile materials [0103] D duct system [0104] T trolley system [0105] R rail system [0106] E electronic device for detecting that a spinning mill has a fault and for estimating a source of the fault [0107] cI configuration information of the spinning mill [0108] kI knowledge-based information of the spinning mill [0109] dF detected fault