METHOD FOR COMPUTER-ASSISTED ADJUSTMENT OF A CONFIGURATION FOR DIFFERENT TEXTILE PRODUCTIONS

Description

[0001] The present invention relates to a method for computer-aided adaptation of a configuration for different textile production processes. The invention further relates to a device, a system, and a computer program for this purpose.

[0002] It is known from the prior art that textile production involves the use of numerous machines, which must be configured to process a specific textile material. The properties of the resulting products depend on a multitude of factors. Consequently, no single configuration can be found that delivers the same production quality across different production runs. Factors such as hardware wear, ambient climate, fiber blends, and material quality vary for each production run, meaning that a consistent configuration can lead to different production results.

[0003] It is therefore also known that the production configurations are determined manually and entered into the respective machine. However, this is time-consuming, and the degree of optimization of each configuration may depend heavily on the experience of the operator performing the configuration.

[0004] German patent application DE 10 2006 014 475 A1 describes how a suitable configuration can be found in off-site trials. However, since this leads to the production of a significant amount of rejects, a method is proposed in which a textile machine is controlled by a neural network. This is intended to avoid the production of rejects. The application of this method is limited to the production of a single textile machine.

[0005] Other generic processes are known from WO 2020 / 100092 A1 and EP 3 760 772 A1.

[0006] Furthermore, the document IT 2019 0000 9465 A1 discloses a method for controlling an operating parameter of an industrial machine, for example a textile machine, in order to change its operating state.

[0007] It is therefore an object of the present invention to at least partially overcome the disadvantages described above. In particular, it is an object of the present invention to propose an improved solution for adapting the configuration for different production processes.

[0008] The foregoing problem is solved by a method, a device, a system, and a computer program having the features of the corresponding independent claims. Further features and details of the invention will become apparent from the respective dependent claims, the description, and the drawings. Features and details described in connection with the method according to the invention naturally also apply in connection with the device, the system, and the computer program according to the invention, and vice versa, so that the disclosure of the individual aspects of the invention always makes, or can make, reciprocal references.

[0009] The problem is solved in particular by a method for computer-aided adaptation of a configuration for different textile production processes, in which, in at least one production step, a textile material, such as fibers, is processed by at least one machine based on a production specification, preferably to manufacture a textile product such as a thread. The method can be implemented as a computer-implemented process or only partially by a computer. The latter can mean that process steps are also provided which are carried out by the at least one machine.

[0010] Furthermore, it may be stipulated that the production specification is varied for different production runs. The production specification is, for example, a definition of the textile product, i.e., the production output of the respective production run. The production specification can vary for different production runs, even though the production runs are still carried out using the same machine(s). It is possible that, depending on the production specification, the textile material varies and / or at least one machine is operated differently. It is also conceivable that environmental factors change during production. The production specification can also define a composition of the textile material, e.g., a fiber blend.

[0011] The process comprises the following steps, preferably for automated production preparation of the respective textile production with the production specification varied for this purpose, wherein the steps are preferably carried out one after the other in the specified order and / or repeatedly: Initiating the execution of at least one or more production experiments, wherein in each production experiment at least one production step is configured with a setting and carried out based on the production specification, determining an effect of the configuration based on the setting in the respective production experiment, adjusting the setting based on the determined effect in order to determine a setting optimized for the production specification.

[0012] It may be planned that after the execution of the first of several production experiments, the settings are adjusted for each subsequent production experiment. Thus, at least one production step in each subsequent production experiment is configured with the appropriate setting. Only after the execution of the last production experiment can the most recently adjusted setting be determined as the optimized setting. This is described below using at least three production experiments as examples, with the following steps being executed sequentially in the given order: First, the execution of the first of several production experiments is initiated. In this experiment, at least one production step is configured with an initial, possibly predefined, setting and carried out based on the production specification. The effect of the configuration based on the initial setting is then determined for the first production experiment. The initial setting is then adjusted based on the determined effect. Next, the execution of a second of several production experiments is initiated. In this experiment, at least one production step is configured with the previously adjusted setting and carried out based on the production specification. The effect of the configuration based on the previously adjusted setting is then determined for the second production experiment. The setting is then adjusted again based on the determined effect.The execution of one of the last of several production experiments is then initiated, in which at least one production step is configured with the previously adjusted setting and carried out based on the production specification. Subsequently, the effect of the configuration based on the previously adjusted setting is determined in the last production experiment. The setting is then adjusted again based on the determined effect. The adjusted setting is then determined as the optimized setting.

[0013] Furthermore, the method can include the following step, preferably for the (automated) initiation of the respective textile production with the production specification varied for this purpose, which is preferably carried out after the setting optimized for the production specification has been determined: Initiating the execution of textile production, in which at least one production step is configured with the optimized setting and carried out based on the production specification.

[0014] It is possible that initiating the implementation of textile production includes the following step: Automated execution of the configuration of at least one machine with the optimized settings.

[0015] Furthermore, at least one machine learning method can be used to perform the adjustment based on the determined effect, preferably being implemented as at least one kernel method. This enables the automatic improvement or creation of production settings by artificial intelligence. The setting adjusted by the at least one machine learning method can then be used for another production experiment to configure the at least one machine or ultimately determined as the optimized setting. In this way, the at least one machine learning method can control the production experiments and ultimately the production process.

[0016] It is possible for the production experiments to be initiated and / or carried out fully automatically. Accordingly, at least one machine learning method can independently control the production experiments, e.g., spinning trials, by adjusting the settings. It is also possible for the impact to be determined automatically. The impact is represented, for example, as at least one measurement result from a detection device such as a sensor. Determining the impact can include, for example, an evaluation of the production experiment and / or the textile material processed and / or the production result and / or other factors such as energy consumption.Since the production experiment can be carried out with the same at least one machine, i.e., the same hardware, as the production, in particular also with the same textile material such as a fiber composite according to the production specification and / or the same environmental conditions such as a climate in the vicinity of the at least one machine, the at least one machine learning method can determine the best setting for the specific production.

[0017] The minimum number of machines can be either a rotor spinning machine or an air spinning machine. It is also possible that the minimum number of machines comprises only a single machine, so that during production, processing is carried out, for example, solely by a machine such as a rotor spinning machine or an air spinning machine.

[0018] It is possible that the respective production run is intended only for the production of a limited, smaller quantity according to the respective production target. Therefore, frequent variations of the production target may be planned, e.g., after a maximum of one week, or even just a few days or hours. It is also possible that the production preparation steps are repeated after each variation of the production target. Furthermore, the respective production run may be limited with regard to the quantity produced and / or production time, so that the respective production run is stopped after a predetermined maximum limit of quantity and / or production time is exceeded, in order to repeat the production preparation steps. This has the advantage that if the production target and / or influencing factors such as the age of at least one machine change, the settings can be re-optimized.

[0019] Furthermore, the invention may provide that the setting comprises at least one or more setting values ​​for the at least one machine in the form of at least one textile machine, preferably a spinning and / or winding machine, in order to control the processing of the textile material, preferably the processing of fibers into yarn (also referred to as thread) and / or the winding and / or rewinding of the yarn, in the at least one production step. For this purpose, the setting value or values ​​can be transmitted to a machine, for example, via a data connection such as a network, in order to configure the machine and thus also the production step carried out by the machine.

[0020] It is also conceivable that the at least one or more setting values ​​comprise at least one of the following: a take-off speed, a rotational speed of a machine rotor, a vacuum, a tension of the textile material and / or a production result during winding, a type and / or selection of the textile material, a winding speed, in particular of a yarn winding device, a setting for a cleaner, in particular a limit value for cutting the textile material by the cleaner. The cleaner can preferably be designed as a sensor or include a sensor that performs a measurement for quality control of the yarn. Furthermore, it can be provided that the cleaner compares the measured values ​​with at least one limit value in order to detect a limit violation if the limit value is exceeded and / or not reached.This makes it possible to determine the sufficient quality of the yarn based on the comparison. In the event of a limit being exceeded, the cleaner can send a signal to initiate the cutting out of the defective piece of yarn.

[0021] Optionally, determining the effect can be carried out by receiving at least one measurement result from at least one measurement device. The measurement device can be part of a machine or arranged on a machine. The measurement device can be designed to detect a measured variable that is influenced by the setting, i.e., represents an effect of the setting. Accordingly, the measurement result can result from a detection, in particular measurement and / or analysis, on the machine and / or on the textile material and / or in the machine's environment and / or on a production output, preferably a yarn, from the respective production experiment.

[0022] The at least one detection device can include a capacitive sensor to measure the capacity of a volume through which the textile material and / or the production product flows, such as a thread. In this way, the quantity flowing through can be determined. It can also be used to determine whether and / or to what extent thin spots occur on the thread. Alternatively or additionally, the at least one detection device can include an optical sensor. In this case, a light-emitting diode of the sensor can illuminate the textile material and / or the production product, such as the thread, and the sensor measures the incoming light to, for example, assess the diameter of the thread. It is also possible for the at least one detection device to include an analytical device that performs a detailed analysis of the textile material and / or the production product.Furthermore, at least one detection device can be configured to perform length and / or weight measurements in order to determine a length-to-weight ratio. The at least one detection device can also include at least one optical sensor to detect foreign fibers. In addition, the at least one detection device can be configured to record production information, such as how often the yarn broke during manufacturing and / or had to be re-spun. This production information can also be provided as at least one detection result. It is also conceivable that the at least one detection device is configured to determine the twist and / or tensile force of the yarn.

[0023] Preferably, the production experiments can be carried out sequentially, with the effect being determined for at least one of the experiments and the settings being adjusted based on this effect. In each subsequent experiment, the configuration is then applied using the adjusted settings. This allows the production experiments to build upon one another and be controlled by the adjustment step. Optionally, the configuration can be fully automated, for example, by transmitting the settings to the relevant machine. Alternatively, the machine itself can collect the necessary data to perform the configuration.

[0024] For example, it may be planned that the several production experiments are carried out sequentially, with the settings being varied in each experiment and the resulting effect being used as input for the machine learning method to optimize the settings for specific production, preferably based on at least one target specification. Since the same production specification is used in the multiple production experiments, which are conducted to prepare for a particular production run, and this same specification is also used for the subsequent production run, this can be described as production-specific optimization. The at least one target specification can also include an optimization goal, which is, for example, predefined.Advantageously, at least one objective can include at least one of the following: maximizing energy efficiency, preferably generating negative and / or positive pressure in the machine; maximizing the quality of the production result, in particular yarn; maximizing production speed; minimizing cleaner cuts; optimizing the machine for efficiency; and verifying effects in effect yarn.

[0025] Preferably, the multiple production experiments can be carried out sequentially, with the final production experiment including determining the effect of the respective production during the production preparation phase. The settings are then adjusted based on this determined effect, and the adjusted setting is used as the optimized setting. In this way, the configuration adjustment can be completed.

[0026] Another possibility is that in each production experiment and production run, the textile material is processed to produce a finished product, particularly yarn. The finished product of each experiment may, at least partially or entirely, be rejects and thus differ from the final product of the production run. In other words, the production experiments may intentionally produce rejects, i.e., finished products like yarn that must be discarded. This can be accepted in order to find optimal production settings. Advantageously, the rejects, i.e., the finished product of the experiments, may comprise at most 1% or at most 0.1% of the final product of the production run.

[0027] It is also advantageous if the textile production processes are carried out using at least one machine, but differ with regard to the production specifications, preferably a production result and / or a requirement for the production result and / or a parameterization and / or recipe for the production result and / or the textile material and / or a parameter of the textile material, preferably a fiber count and / or fiber thickness and / or fiber blend. Furthermore, a predefined, initial setting can be used for the production preparation of each production run, which is only varied after the first production experiment has been carried out. Preferably, the optimized setting determined for subsequent production runs is disregarded.This allows optimization to be limited to specific productions. This ensures that settings can be optimized for even very different productions.

[0028] Optionally, it is conceivable that at least one or exactly one machine learning method is implemented as a kernel method, and preferably, that kernel method is implemented as one of the following: Support Vector Machine, Gaussian Process, Kernel PCA, Kernel Perceptron, Canonical Correlation Analysis, Ridge Regression, Spectral Clustering, Linear Adaptive Filters. The listed techniques have the advantage of enabling a fast and precise estimation of error sources. In particular, the listed techniques and algorithms have the advantage of being able to perform robust optimization even with incomplete information. This results in fewer iterations of experiments being required to obtain optimal results.

[0029] Furthermore, it is conceivable that a production-specific portion of training data is provided for training the at least one machine learning method, which is determined exclusively from the at least one or more production experiments of the production preparation for the respective production. Preferably, the training data for the productions comprises identically defined initial training data, in particular a predefined initial setting, and additionally, training data that varies based on the determined effect. This ensures that the settings can be optimized even for very different productions.

[0030] Furthermore, within the scope of the invention, it is optionally possible for at least one or exactly one machine learning method to be implemented as a Gaussian process. This makes it possible to optimize settings for many different productions even with limited training data.

[0031] A further advantage can be achieved within the scope of the invention if the steps for automated production preparation and / or production initiation are carried out fully automatically and / or the method is at least partially or completely computer-implemented and preferably executed in real time. This allows the configuration to be adjusted even without the intervention of an experienced user.

[0032] Furthermore, within the scope of the invention, it is conceivable that in the respective production experiment, at least one production step is carried out in the form of at least or exactly one spinning trial, in which a thread is spun based on the setting, in order to optimize the setting for the respective production based on the production experiment by adjusting it.

[0033] The at least one machine, in particular a textile machine, advantageously comprises at least one of the following: spinning machine, textile machine for fiber preparation, textile machine for spinning preparation, ring spinning machine, compact spinning machine, rotor spinning machine, air spinning machine, automation machine, winding machine, texturing machine. A textile machine can also have a plurality of adjacent workstations. The workstations can be, for example, spinning stations or the like. It is also possible for the workstations to be configured as winding stations, where a yarn is wound onto a spool, e.g., a cross-wound spool. For example, on a spinning machine, a yarn is produced from a pre-prepared fiber composite, which, after leaving a spinning unit, passes—viewed in the direction of yarn travel—successively through a take-up device and a yarn storage unit and is finally wound onto the yarn winding device, i.e.,the winding point.

[0034] To monitor production, at least one detection device, e.g., at least one sensor, can be used. The detection device can determine at least one result, e.g., by detecting the textile material and / or the production result on the textile machine and / or at least one other machine used for production preparation and / or production. The at least one result is, for example, expressed as a measured value or the like. The detection by the device can be performed repeatedly during production preparation and / or production and can thus also be described as monitoring production preparation and / or production. For example, it is known that to monitor the yarn path at the individual workstations of a textile machine, in the area of ​​the yarn path, i.e.,Using the spinning machine as an example, a so-called yarn monitor is positioned between the spinning unit and the yarn winding device. The yarn monitor can be a yarn sensor connected to the workstation's control system, which detects both unintentional yarn breakage and defects in the spun yarn. In both cases, the respective work process can be interrupted via the workstation's control system. If the yarn is damaged, a cleaning cut can be performed to remove the damaged section. If the yarn is broken, a section of the yarn end can be removed. In both cases, the yarn ends can then be joined. On a spinning machine, for example, the yarn end coming from the yarn winding device is then fed to the spinning unit, where a new winding process is carried out, after which the spinning process can resume.During the operation of at least one textile machine and / or at least one machine used for production preparation and / or production, i.e., during production or production preparation, at least one measurement result can be determined by at least one measurement device. Furthermore, connection points can arise at the points where the spinning process is repeated, which can also be the target of optimization.

[0035] The at least one data acquisition result can be at least one of the following: a physical parameter and / or a chemical parameter of the at least one textile machine, such as rotational speed, power consumption, or a chemical property of lubricants. Furthermore, the at least one data acquisition result can include at least one physical parameter and / or a chemical parameter of the textile material, such as thickness, density, a chemical property of the fibers, and the like. The acquisition of parameters from textile machines and / or textile material can include taking samples from the textile machines and / or the textile material and examining the samples in laboratories, in particular to determine the mechanical / physical and / or chemical properties of the samples.

[0036] Production preparation can utilize part or the same production line and / or at least one textile machine as production. Therefore, characteristics described in connection with production also apply to production preparation. Production, and thus production preparation, can comprise one or more production steps. Particularly in the processing of fibers such as natural and synthetic fibers and their blends into yarns, the production steps can include at least one of the following: converting the fibers into carded slivers, drawing, spinning, automatic transport from a preceding textile machine to a subsequent one, rewinding, texturing, and winding. To achieve a desired production capacity along a serial production line, textile machines or components of textile machines can be arranged in parallel.Thus, a preceding production step can be followed serially by a subsequent production step, whereby the preceding production step may require fewer or more parallel textile machines or components for processing textile material than the subsequent production step. For example, the fiber preparation step, which yields ribbons, may require fewer parallel textile machines or components than the step of spinning ribbons into yarns.

[0037] It is possible that at least one data acquisition result includes at least one of the following: machine power consumption, lubricant quality, textile material diameter, textile material fiber density, a result of a laboratory analysis of the textile material, fiber moisture, micronaire, fiber length, thread path, yarn defects, fiber length uniformity, fiber strength, fiber nubs, fiber maturity, fiber color, fiber waste, number of spools / rovings per strand, spool / roving weight variation, spool / roving uniformity, thick spots, thin spots, degree of twist, yarn strength, yarn elongation properties, strength, twist and number, yarn hairiness, yarn abrasion resistance, wear properties, yarn color, environmental parameters of the machine, in particular with regard to climatic conditions, frequency of breakage and / or yarn defects of the textile material and / or production result, in particular yarn, and others.The data collection result can be obtained, for example, by taking samples of lubricants to determine their quality. Similarly, the data collection result for textile machinery and / or textile materials can be based on external information sources, such as quality data from a textile material supplier (e.g., quality information for raw cotton), type designation / product number and / or material properties, and / or expiration dates of consumables such as lubricants. Environmental parameters, particularly climatic conditions such as temperature, humidity, solar radiation, and sun position, preferably measured or based on external information sources and / or date and / or time, can also be used as data collection results. These climatic conditions can be found in a raw material warehouse, a warehouse for intermediate products, and / or a warehouse for finished products.For example, parameters for climatic conditions can also be recorded in at least one area of ​​a spinning mill.

[0038] The invention also relates to a data processing device comprising means for carrying out the steps of the method according to the invention. The device according to the invention thus offers the same advantages as those described in detail with reference to a method according to the invention. The device can be part of at least one or exactly one of the at least one machine.

[0039] The invention also relates to a system for carrying out textile production, comprising: at least one machine for processing a textile material in at least one production step, the device according to the invention for data processing, at least one detection device for determining the effect by detecting at least one detection result on the machine and / or on the textile material and / or in an environment of the machine and / or on a production result.

[0040] The system according to the invention thus offers the same advantages as those described in detail with reference to a method according to the invention.

[0041] The invention also relates to a computer program, in particular a computer program product, comprising instructions which, when executed by a computer, cause the computer to execute the method according to the invention. Thus, the computer program according to the invention offers the same advantages as those described in detail with reference to a method according to the invention.

[0042] The computer can be a data processing device, for example, the device according to the invention, which executes the computer program. The computer can have at least one processor for executing the computer program. A non-volatile data storage device can also be provided in which the computer program is stored and from which the processor can read the computer program for execution. Furthermore, the computer program can also be directly integrated into the machine's software.

[0043] It is also conceivable that the computer includes at least one integrated circuit such as a microprocessor, an application-specific integrated circuit (ASIC), an application-specific standard product (ASSP), a digital signal processor (DSP), a field-programmable gate array (FPGA), or the like. The computer may also have at least one interface for data exchange, such as an Ethernet interface, an interface for LAN (Local Area Network), WLAN (Wireless Local Area Network), a system-on-a-chip (SoC), or another wireless interface such as Bluetooth or near-field communication (NFC). Furthermore, the computer may be implemented as one or more control units, i.e., also as a system of control units. The computer may, for example, also be intended to be located in a cloud and / or as a server to provide data processing for a local application via the interface.It is also possible that the computer is designed as a mobile device, such as a smartphone.

[0044] The invention may also relate to a computer-readable storage medium comprising the computer program according to the invention. The storage medium is, for example, designed as a data storage device such as a hard drive and / or non-volatile memory and / or a memory card. The storage medium may, for example, be integrated into the computer.

[0045] Further advantages, features, and details of the invention will become apparent from the following description, in which exemplary embodiments of the invention are described in detail with reference to the drawings. The features mentioned in the claims and in the description can each be essential to the invention individually or in any combination. The drawings show: Fig. 1 schematically shows an embodiment of the method according to the invention, Fig. 2 schematically shows an embodiment of a system and device according to the invention as well as a computer program according to the invention.

[0046] In the following figures, identical reference numerals are used for the same technical features even for different embodiments.

[0047] In Fig. 1Figure 100 illustrates an embodiment of a method 100 according to the invention for the computer-aided adaptation of a configuration for different textile productions 210. It is shown by way of example, using a timeline t, that several productions 210 can be carried out sequentially, wherein a production specification 230 is varied for the different productions 210 (exemplary varied production specifications 230 are schematically represented on the timeline t by an arrow). Furthermore, in at least one production step 220 of each production 210, a textile material, such as fibers, is processed by at least one machine 10 to manufacture a textile product, such as a thread, based on the production specification 230. The processing of the textile material and the machine 10 are shown in Figure 10. Fig. 2 illustrated with further details.

[0048] Furthermore, in Fig. 1It has been shown that process steps 101, 102, 103, and 104 are provided for the automated production preparation and initiation of the respective textile production 210 with the varied production specification 230. According to a first process step 101, the execution of at least one or more production experiments 215 is initiated. In each production experiment 215, at least one production step 220 can be configured with a setting 240 and carried out based on the production specification 230. Subsequently, according to a second process step 102, an effect 250 of the configuration based on setting 240 is determined for the respective production experiment 215. According to a third process step 103, the setting 240 is adjusted based on the determined effect 250 in order to determine a setting 240 optimized for the production specification 230.Furthermore, a dashed arrow indicates that process steps 101 to 103 can be performed iteratively. This means that after conducting the first of the production experiments 215, the effect 250 can first be determined in order to adjust setting 240 accordingly. This adjusted setting 240 can then be used to perform a second of the production experiments 215. The effect 250 can again be determined in order to adjust setting 240 once more and use it for a third of the production experiments 215, in order to configure at least one production step 220. This procedure can optionally be repeated for further production experiments 215.The several production experiments 215 are thus carried out one after the other, between the production experiments 215 the determination 102 of the effect 250 and the adjustment 103 of the setting 240 is carried out, whereby for a last of the production experiments 215 of the production preparation of the respective production 210 the determination 102 of the effect 250 and subsequently the adjustment 103 of the setting 240 on the basis of the determined effect 250 is finally carried out in order to use the setting 240 adjusted in this way as the optimized setting 240.

[0049] With the final iteration of step 103, production preparation is complete, and the determined optimized setting 240 can be used for a fourth process step 104 when initiating textile production 210. During textile production 210, at least one production step 220 is also carried out based on the production specification 230, but configured with the optimized setting 240.

[0050] Furthermore, it is illustrated that at least or exactly one machine learning method 300 can be used to perform the adaptation 103 based on the determined impact 250. Preferably, the machine learning method 300 is implemented as a kernel method 300.

[0051] In Fig. 2The inventive method 100 is illustrated by the example of a specific textile machine 10. The textile machine 10 can be provided as part of a system 1 in which a device 20 for data processing, a computer program 30, and at least one detection device 15 for determining 102 the effect 250 by detecting a detection result on the machine 10 and / or on the textile material 2 and / or in an environment of the machine 10 and / or on a production result 3 are also provided. In the specific example, the textile material 2 can comprise fibers, i.e., in particular a fiber composite, and the production result 3 can comprise a thread. In this context, the Fig. 1In the respective production experiment 215 and the respective production 210, the textile material 2 is processed to produce the production result 3, whereby the production result 3 of the respective production experiment 215 constitutes at least a partial or complete reject and thus differs from the production result 3 of the respective production 210. The same process can be used in both production experiments 215 and production 210. Fig. 2The machine 10 shown can be used to carry out the optimization under the conditions actually present in production 210. Furthermore, the other textile productions 210 can carry out the processing with the same machine 10, but differ with regard to the production specification 230, preferably the production result 3 and / or a requirement for the production result 3 and / or a parameterization and / or recipe for the production result 3 and / or the textile material 2 and / or a parameter of the textile material 2, preferably a fiber count and / or fiber thickness.

[0052] It is specifically intended that a machine learning method 300, referred to as AI for short, is initially given a preliminary setting 240 for production 210. Using the AI, adjusted settings 240 can then be transferred to a connected spinning or winding station, thereby spinning and / or winding a thread 3. Detection devices 15, in particular sensors, can then record the result of these production experiments 215, e.g., the thread 3, as the effect 250 and transmit it back to the AI. From this data set, the AI ​​can derive a new setting 240 and perform another production experiment 215. Over time, the AI ​​would continuously improve the settings. The AI's objectives can be various criteria, such as increasing yarn quality, minimizing purging, improving energy efficiency, or increasing production speed.

[0053] It is possible that the process steps or the computer program 30 for carrying out the process steps are executed in real time. This allows for a fully automated real-time application of the AI ​​described above. In this way, the AI ​​can be used—for example, by the user—before each new production run 210 to determine the optimized setting 240. This means that the setting 240, with its associated settings, is optimized for a specific production run 210 and is no longer needed after the production run 210 is completed (i.e., possibly after just a few hours). Accordingly, the training time for the AI ​​is relatively short, and only a small amount of training data from the production experiments 215 is available.Thus, an AI-driven optimization process is provided, which is suitable for determining the optimized setting 240 for each production preparation in a very short time using only a small amount of training data. If this is done on the user side, it is further advantageous if the necessary production experiments 215 are carried out as automatically as possible on machine 10. It is also possible for samples 250 to be automatically transferred from the production experiments 215 to a laboratory device to determine the effect.

[0054] Various methods are known for implementing AI, including the conventional use of neural networks. However, due to the limited availability of training data, kernel methods are preferred according to the invention. For example, a Gaussian process can be used as the AI. In contrast to neural networks, these classical machine learning models offer the possibility of reliably performing optimization even with limited training data. This makes it possible to apply the optimization method even to short production processes and on the user side. It is also possible that the invention does not require the use of a neural network, meaning that at least one machine learning method does not include a single neural network.

[0055] The device 20 can be part of one or more of the at least one machine 10 for textile production, enabling the user to carry out process steps 101-104. In production 210, a large number of textile machines 10 may be used, for example, to convert natural and synthetic fibers and their blends into yarns of a desired type, quantity, and quality. The desired type, quantity, quality, and / or blend can be defined, for example, by the production specification 230 and vary for different production runs 210. Various types of textile machines 10 can be used for production, such as winding machines or spinning machines. These include, in particular, air-jet and rotor spinning machines, which are generally known from the prior art.

[0056] It is also known that measurements are carried out on textile machines 10 and on the processed textile material 2 in order to monitor production. For this purpose, at least one detection device 15 can be used, such as the one described in Fig. 2 the two sensors shown 15. This shows Fig. 2A spinning station 11 is schematically represented, with a thread 3 extending from a spinning unit 21 (e.g., an air-jet or rotor spinning unit) to a thread winding device 13 during the spinning process. A fiber composite 2 fed to the spinning unit 21 is twisted within the spinning unit 21 after passing through a drafting unit (not shown) in the case of an air-jet spinning machine or a disentangling unit in the case of a rotor spinning machine. Downstream of the spinning unit 21 – relative to the thread travel direction 22 – is a take-up device, visualized by take-up rollers 19, which uses the take-up roller pair 19 to draw the thread 3 exiting the spinning unit 21 and transport it in the direction 22 to the thread winding device 13.Downstream of the take-up device is a thread storage unit with a thread storage tube 17 (dashed line), on which a first and a second thread sensor 15 are arranged in the area of ​​a thread entry opening (not shown). The thread storage tube 17 is positioned such that the thread sensors 15 are located in the area of ​​the thread path. The thread sensors 15 thus form a sensor system by means of which the thread path can be detected. In this way, a thread defect and / or a break of the thread 3 can also be detected as a result of the detection. The loop-shaped thread section arranged inside the thread storage tube is generated by suction air directed from the thread entry opening towards the interior of the thread storage tube. This suction air is provided by a vacuum source 18, which is connected to one end of the thread storage tube 17 opposite the thread entry opening.The yarn sensors 15 thus detect the extent to which the loop-shaped yarn section extends within the yarn storage tube. In the event of a yarn breakage or a yarn defect detected by the yarn sensor 15, the spinning process is interrupted by the connected control unit. Subsequently, a defined free yarn end coming from the yarn winding device 13 is created by means of a cutting device 16 and, in a further step, conveyed via return means 14 into the spinning unit 21. Here, the yarn end is spun onto the yarn 3 coming from the spinning unit 21. Furthermore, in . Fig. 2 A drive 12 is shown, which serves to determine the winding speed of the thread winding device 13.

[0057] To perform the adjustment 103 based on the acquisition result provided by the acquisition device 15, the at least one acquisition result and / or the at least one setting value and / or at least one target specification can be passed as input to the at least one machine learning method 300 in order to obtain at least one optimized setting value as output by applying the machine learning method 300. For the application of the machine learning method 300, a data set from the acquisition result can be prepared and used as input for the method 300. For example, to prepare the data set, at least one setting value and at least one target specification can initially be stored in a file. Preferably, a target specification is predefined for each acquisition result, which indicates the value to which or with which goal (e.g.,Maximizing the measurement result, which should be optimized towards at least one initial setting value. Specifically, this could be, for example, increasing energy efficiency, yarn quality, production speed, or a combination thereof. The target for production speed, for instance, defines a duration or a minimization of that duration.

[0058] The subsequent application of method 300 is described below using the example of a Gaussian process model – referred to simply as the Gaussian process – for several setpoint values ​​and acquisition results. Like linear regression models, tree-based models, or perceptron-based models, the Gaussian process model belongs to the methods of machine learning. One advantage is that it is a machine learning model that can be solved analytically. It will be shown below that its use for optimizing setpoint values ​​for textile machines 10 can surprisingly improve production 210.

[0059] During production preparation, the data set can always contain the current settings of setting 240 (which may be adjusted) as well as the dependent and repeatedly determined measurement results according to effect 250. The settings can thus be defined as the target value yn by the function f(xn), where xn represents the individual inputs from x = [x₁, x₂, ..., xφ]T, which can be derived from the measurement results, and en is independent Gaussian noise. y n = f x n + e n .

[0060] Furthermore, the conditional probability of observing yn at f(xn ) can be determined by the normal distribution: p y n f x n = N y n f x n , σ , where σ is the standard deviation of en and σ = σl is a diagonal matrix of size φ × φTo make predictions about y, the marginal probability distribution p(y) can be determined. This probability distribution can be obtained by marginalizing the conditional distribution p(y|f(x)) over the distribution p(f(x)) using the integral: p y = ∫ p y f x − p f x df x .

[0061] The distribution p(f(x)) is defined as a Gaussian distribution with a mean of 0 and a covariance kernel matrix K of size φ × φ : p f x = N f x 0 , K . with K n m = k x n x m .

[0062] This results in: p y = N y 0 , C , where each element in C can be obtained by: C [ n , m ] = k ( xn , xm ) + σδ nm .

[0063] Various covariance kernel functions, such as a constant kernel, a quadratic exponential kernel, or a periodic kernel, can be used for K or k(xn, xm). The quadratic exponential kernel can be computed from pairs of samples (xn, xm) in x: k x n x m = exp − x n − x m 2 / 2 , the samples can be obtained at the beginning from the initially intended setting values.

[0064] To optimize the settings, it may now be necessary to use the existing φ Given inputs x = [x₁, x₂, ... xφ] ∈ T< and known target values ​​sy = [y₁, y₂, ... yφ] ∈ T<, the values ​​of yφ+1 corresponding to a new input xφ+1 are predicted. The new input can correspond to the target specifications. To determine the parameters of p(yφ+1 |y), the distribution p(y') can be used, where y' = [y₁, y₂, ... yφ, yφ+1] ∈ T< is a vector of length φ +1. Thus, p(y') = N(y'|0, C'), where the new covariance matrix C' is of size φ + 1 × φ + 1 with the structure C' = [[ C , k ],... [ k T< , c ]] is, where C = K+σI is the original φ × φ -covariance matrix is, k is a vector of length φ is whose elements are given by: k[n] = k(xn , x φ+1 ), and c is a scalar containing the covariance of x φ+1 with itself: c = k(x φ+1 , x φ+1 )+σ. The subsequent adjustment 103 of setting 240 can be made based on the values ​​predicted in this way.

[0065] The preceding explanation of the embodiments describes the present invention solely by way of examples. Naturally, individual features of the embodiments can be freely combined with one another, provided this is technically feasible, without departing from the scope of the present invention. Reference symbol list

[0066] 1 system 2 textile material, fiber composite 100 Proceedings 3 Production result, thread 101 First procedural step, initiation 102 second procedural step, determining 10 machine 103 third process step, adjusting 11 Workplace, spinning station 104 fourth process step, initiating 12 drive 13 Thread winding device 210 production 14 Recycled materials 215 Production experiment 15 Detection device, thread sensor 220 Production step 16 Separating device 230 Production specification 17 Thread storage tube 18 source of negative pressure 240 Attitude 19 Take-off roller 250 impact 20 device 300 machine learning method, 21 Spinning unit Kernel method 22 grain direction t Time 30 Computer program

Claims

1. A method (100) for computer-assisted adjustment of a configuration for different textile productions (210) with which a production specification (230) is varied for the different productions (210) and with which, in each case in at least one production step (220) based on the production specification (230), a textile material (2) is processed by at least one machine (10), comprising the following steps for automated production preparation and production initiation of the respective textile production (210) with the production specification (230) varied for this purpose: - Initiating (101) a performance of at least one or more production experiments (215), wherein with the respective production experiment (215) the at least one production step (220) is configured with a setting (240) and is performed based on the production specification (230), - Determining (102) an effect (250) of the configuration based on the setting (240) in the particular production experiment (215), - Adjusting (103) the setting (240) on the basis of the determined effect (250), in order to determine a setting (240) optimized to the production specification (230), - Initiating (104) a performance of textile production (210) with which the at least one production step (220) is configured with the optimized setting (240) and is performed based on the production specification (230), wherein at least one machine learning method (300) is used to perform the adjustment (103) on the basis of the determined effect (250), wherein preferably the at least one machine learning method (300) is executed as at least one kernel method (300).

2. Method (100) according to claim 1, characterized in that the setting (240) comprises at least one or more setting values for the at least one machine (10) in the form of at least one textile machine, preferably a spinning and / or winding machine, in order to control the processing of the textile material (2), preferably a processing of fibers into yarn and / or a rewinding and / or winding of the yarn, in the at least one production step (220), wherein preferably the at least one or more setting values comprise at least one of the following: a take-off speed, a rotational speed of a rotor of the machine, a negative pressure, a tension of the textile material (2) and / or a production result (3) during winding, a type and / or selection of the textile material (2), a setting for a clearer, in particular a limit value for a cutting of the textile material (2) by the clearer.

3. The method (100) according to any one of the preceding claims, characterized in that the determination (102) of the effect (250) is performed as receiving at least one detection result from at least one detection apparatus (15), wherein the detection result results from a detection, in particular measurement and / or analysis, on the machine (10) and / or on the textile material (2) and / or in an environment of the machine (10) and / or on a production result (3), preferably a yarn, of the respective production experiment (215), wherein preferably the at least one detection result comprises at least one of the following: a rotational speed of a rotor of the machine (10), a power consumption of the machine (10), a quality of a lubricant, a diameter of the textile material (2), a fiber density of the textile material (2), a result of a laboratory analysis of the textile material (2), environmental parameters of the machine (10), in particular with respect to climatic conditions, a frequency of a breakage and / or a thread defect of the textile material (2) and / or production result (3), in particular thread.

4. The method (100) according to any one of the preceding claims, characterized in that the plurality of production experiments (215) are performed sequentially, wherein, for at least one of the production experiments (215), the determination (102) of the effect (250) and subsequently the adjustment (103) of the setting (240) is performed on the basis of the determined effect (250) in order to perform the configuration with the setting adjusted thereby upon a respective subsequent experiment of the production experiments (215).

5. The method (100) according to any one of the preceding claims, characterized in that the plurality of production experiments (215) are performed sequentially, wherein, in each of the production experiments (215), the setting (240) is varied by the adjustment (103) and the effect (250) determined thereby is used as input for the machine learning method (300) in order to optimize the setting (240) in a production-specific manner, preferably based on at least one target specification, wherein the at least one target specification preferably comprises at least one of the following: an energy efficiency, preferably a negative pressure and / or positive pressure generation at the machine, a quality of the production result (3), in particular yarn, a production speed.

6. The method (100) according to any one of the preceding claims, characterized in that the plurality of production experiments (215) are performed sequentially, wherein, for a last experiment of the production experiments (215) of the production preparation of the respective production (210), the determination (102) of the effect (250) and subsequently the adjustment (103) of the setting (240) is performed on the basis of the determined effect (250) in order to use the setting (240) adjusted thereby as the optimized setting (240).

7. The method (100) according to any one of the preceding claims, characterized in that with the respective production experiment (215) and the respective production (210) the textile material (2) is processed to produce a production result (3), in particular a yarn, wherein the production result (3) of the respective production experiment (215) at least partially or completely forms a reject and thus differs from the production result (3) of the respective production (210).

8. The method (100) according to any one of the preceding claims, characterized in that the textile productions (210) perform the processing with the same at least one machine (10), but differ with respect to the production specification (230), preferably a production result (3) and / or a requirement for the production result (3) and / or a parameterization and / or recipe for the production result (3) and / or the textile material (2) and / or a parameter of the textile material (2), preferably a fiber count and / or fiber strength, wherein a predefined, initial setting (240) is used as the setting (240), in each case for the carrying out of a first experiment of the production experiments (215) of the production preparation of the respective production (210), which is only varied after the carrying out of the first production experiment (215), wherein preferably the respectively determined optimized setting (240) of the further productions (210) is not taken into account.

9. The method (100) according to any one of the preceding claims, characterized in that the at least or exactly one machine learning method (300) is executed as a kernel method (300), and preferably the kernel method (300) is executed as one of the following: support vector machine, Gaussian process, kernel PCA, kernel perceptron, canonical correlation analysis, ridge regression, spectral clustering, linear adaptive filters, and / or that a production-specific portion of training data is provided for the training of the at least one machine learning method (300), which is determined exclusively from the at least one or more production experiments (215) of the production preparation of the respective production (210), wherein preferably the training data for the productions (210) comprises equally defined initial training data, in particular a predefined, initial setting (240), and additionally training data varying on the basis of the determined effect (250).

10. The method (100) according to any one of the preceding claims, characterized in that the at least or exactly one machine learning method (300) is executed as a Gaussian process.

11. The method (100) according to any one of the preceding claims, characterized in that the steps for automated production preparation and production initiation are performed fully automatically and / or the method (100) is at least partially or completely computer-implemented and is preferably executed in real time.

12. The method (100) according to any one of the preceding claims, characterized in that with the respective production experiment (215) the at least one production step (220) is performed in the form of at least or exactly one spinning test with which a thread is spun based on the setting (240) in order to optimize the setting for the respective production (210) based on the production experiment (215) by adjusting (103) it.

13. An apparatus (20) for data processing comprising means for executing the steps of the method (100) of any one of the preceding claims.

14. A system (1) for carrying out textile productions (210), comprising: - at least one machine (10) for processing a textile material (2) in at least one production step (220), - the apparatus (20) for data processing according to claim 13, - at least one detection apparatus (15) for determining (102) the effect (250) by detecting at least one detection result on the machine (10) and / or on the textile material (2) and / or in an environment of the machine (10) and / or on a production result (3).

15. A computer program (30) comprising instructions that, when the computer program (30) is executed by a computer, cause the computer to execute the steps for automated production preparation and production initiation of the method (100) according to any one of the claims 1 to 12.