RU2798082C1 - Preliminary assessment of the quality of a fibre optic cable based on machine learning during the manufacturing process - Google Patents

Abstract

FIELD: fibre optics.

SUBSTANCE: invention relates to the manufacture of a fibre optic cable, namely to the provision of a means for assessing the quality of a fibre optic cable during the manufacturing process. The claimed method for quality control of a loose-laid fibre optic cable during manufacture on a secondary coating line includes the steps of: storing a trained machine learning algorithm in a machine learning database to calculate the expected values of one or more fibre optic cable quality metrics with loose laying of fibres in the tube, where the fibre optic cable is produced on the recoating line, based on the values of one or more process parameters of the recoating line; one or more values of one or more process parameters of the secondary coating line are controlled by means of a computer system during operation of the secondary coating line; one or more expected values of one or more quality metrics are calculated by the computing system during real-time control using the trained machine learning algorithm, while the controlled values of one or more process parameters are used as an input parameter of the trained machine learning algorithm; calculating in real time monitoring one or more optimum values of one or more control parameters of the recoating line to improve the quality of the manufactured fibre optic cable with loose fibres in the tube, as determined by one or more expected values of one or more quality metrics; and initiating, during real-time monitoring, adjustment of one or more control parameters of the secondary coating line to match one or more optimal values of one or more control parameters.

EFFECT: increased quality control of the manufactured fibre optic cable.

19 cl, 7 dwg

Description

FIELD OF TECHNOLOGY

The present invention relates to the manufacture of an optical fiber cable, namely to providing a means for evaluating the quality of an optical fiber cable during the manufacturing process.

BACKGROUND OF THE INVENTION

Loose-laid fiber optic cable is a type of fiber optic cable in which glass fibers for transmitting optical signals are loosely encased in a semi-rigid protective sheath or tube (called loose-laid tube). The quality of the loose fiber optic cable produced on the so-called recoating line is usually controlled by periodically measuring various key properties of the newly manufactured fiber optic cable off the line. For example, the key properties mentioned to be measured may include excess fiber length, tube shrinkage, and fiber attenuation. Obviously, such measurements outside the line of an already manufactured fiber optic cable require additional resources, and can also take quite a long time. Moreover, any deterioration in the quality of the manufactured fiber optic cable with loose fibers in the tube can be detected only after a considerable time. At least some of the key properties can also be measured during manufacture, although such in-line measurements are often inaccurate compared to off-line measurements. Thus, there is a need for a solution that would enable quality control of fiber optic cables in a more automated way during production without sacrificing accuracy.

BRIEF DESCRIPTION OF THE INVENTION

According to an aspect, the subject matter of the invention is provided in independent claims. Embodiments are presented in dependent claims.

One or more embodiments are set forth in more detail in the accompanying drawings and description below. Other features will become apparent from the description and drawings, as well as from the claims.

BRIEF DESCRIPTION OF THE DRAWINGS

Hereinafter the invention will be described in more detail with the help of preferred embodiments and with reference to the accompanying drawings, where:

On FIG. 1 shows a system in accordance with embodiments;

On FIG. 2-5, 6A and 6B show processes in accordance with embodiments; And

On FIG. 7 shows a device according to embodiments.

DETAILED DESCRIPTION OF EMBODIMENTS

Loose-lay fiber optic cable is a type of fiber optic cable specifically designed for use in harsh environments. In loose tube fiber optic cables, the glass fibers for optical signal transmission are loosely bundled in a semi-rigid protective sheath or tube (called loose tube or loose buffer). The space within said loose-laid tube that is not occupied by glass fibers is typically filled with a filling compound such as a waterproof gel, electrolyte or waterproof yarn.

In many cases, the most critical phase in the manufacture of loose-laid fiber optic cables is the recoating phase carried out in the recoating line. In the recoating line, the glass fibers (and possibly the filling compound of the tube) are inserted into a loose-laid tube (usually made of plastic). Initially, when the loose-laid tube is extruded during the secondary coating phase, it has a very high temperature. As the free-laid tube cools, it shrinks or shrinks, which results in a change in the length of the glass fibers in the tube and in the tube itself (in other words, an increase in the so-called excess fiber length). To eliminate this effect, a so-called compression track is used, which affects the speed of transfer of the tube along the line (relative to the speed of transfer of fibers along the line) so as to compensate for the shrinkage of the loose-laid tube. By using a compression track, very high line speeds (eg up to 1000 meters per minute) can be achieved while maintaining almost zero excess fiber length.

The quality of the fiber optic cable produced in the secondary coating line is usually controlled by periodically taking measurements of various key properties of the newly manufactured fiber optic cable. For example, the key properties mentioned to be measured may include excess fiber length, tube shrinkage, and fiber attenuation. Obviously, such measurements outside the line of an already manufactured fiber optic cable require additional resources, and can also take quite a long time. For these reasons, the above measurements can only be carried out from time to time, and therefore, any deterioration in the quality of the manufactured fiber optic cable can only be detected after a considerable time.

The purpose of the following embodiments is to solve or at least mitigate at least some of the mentioned problems associated with the quality control of the manufactured fiber optic cable with loose fibers in the tube.

In the following, various embodiments of the invention will be described in detail. Said embodiments of the invention are based on the use of a machine learning algorithm to pre-assess the quality of an optical fiber during manufacture on a recoating line. To facilitate a detailed explanation of the embodiments, first, machine learning algorithms that may be used in connection with the embodiments are explained in detail.

The machine learning algorithm according to the embodiments may be based on one or more neural networks. Neural networks (or more precisely, artificial neural networks) are computing systems composed of highly interconnected "neurons" capable of processing information through their response in a dynamic state to external input signals. In other words, an artificial neural network is a group of interconnected nodes (or "neurons"), where each connection between nodes is associated with a weight (i.e. weight coefficient), the value of which affects the strength of the signal in said connection, and, therefore, also on the overall performance of the neural network. Typically, the offset component is also added to the total weighted sum of the input signals at the node. Training a neural network typically involves adjusting said weights and biases so that a known output matches a given known input.

One or more neural networks employed in embodiments may comprise one or more feed-forward neural networks, one or more recurrent neural networks, and/or one or more self-organizing maps (SOMs). In addition, one or more feed-forward neural networks may comprise one or more layered perceptron networks and/or one or more convolutional neural networks. One or more recurrent neural networks may contain one or more networks with long-term and short-term memory and/or one or more recurrent convolutional neural networks.

An example of a feed-forward neural network that may be used in embodiments is the "multilayer perceptron" model (although a simple perceptron may also be used in some embodiments). A single layer perceptron can be used to train linearly separable functions, but cannot be used to perform complex tasks such as learning a non-linear decision boundary in classification. On the other hand, a multilayer perceptron that uses two or more layers of perceptrons can be used to train complex features and highly non-linear decision boundaries. A multilayer perceptron is the basic form of a feedforward neural network and typically consists of an input layer, one or more hidden layers, and an output layer. This network uses forward passes and backpropagation to train weights and biases. Forward passes (input to output) calculate the output signals, while error backpropagation calculates the necessary updates to the weights and biases based on the error in the output layer.

Feedforward neural networks are not capable of storing any information because they do not have contours. Recurrent neural networks (RNNs), in contrast, have circuits that allow information to be stored. An example of a recurrent neural network that may be employed in embodiments is a long-term and short-term memory (LSTM) network, which is a special type of recurrent neural network specialized in distant dependency learning. One LSTM cell consists of three gates (input, output and forget) and a memory cell. Gates act as information regulators and help LSTM cells remove old information or add new information. The extent to which existing memory is forgotten is controlled by the forget gate.

A self-organizing map is a type of neural network that is trained using unsupervised learning to obtain a low-dimensional discretized representation of the input training sample space (called a map). In other words, self-organization can be used to create a low-dimensional representation or visualization of high-dimensional data.

Additionally or alternatively, the machine learning algorithm according to the embodiments may be based in whole or in part on a Bayesian classifier, i. e. classifier based on Bayesian probability. The Bayesian classifier used may be, for example, a Naive Bayesian classifier.

The system to which the embodiments can be applied is shown in FIG. 1. In FIG. 1 is a simplified system showing only some of the elements and functional entities. The connections shown in Fig. 1 are logical connections; actual physical connections may vary. One skilled in the art will appreciate that the systems also contain other functions and structures. Fig. 1 is essentially to be considered largely schematic.

The system shown in Fig. 1 comprises two distinct parts: a recoating line 101 for manufacturing a loose-laid fiber optic cable and a control system 116 for controlling, monitoring and/or controlling said recoating line 101.

The recoating line 101 is used to extrude the loose-laid tube, feed the fiber optics (typically glass fibers) into the loose-laid tube in a controlled manner, cool and wind the resulting loose-laid fiber optic cable 104 into the tube. Coating line 101 includes at least one or more drives 102, a fiber feeder 103, an extruder 105, a compression track 106, a middle hauler 107, an end hauler 108, and a winder 109.

The fiber feeder 103 is used to hold the fiber (ie glass fiber) and to feed one or more fibers into the extruder 105 and down the overcoat line 101 in a controlled manner, preferably at high speed. The tension of the fiber can be adjusted, for example, with a pneumatic roller. The fiber feeder 103 may include multiple positions for fiber spools.

An extruder 105 (or extruder) is used to extrude a (plastic) material to form a loose-laid tube structure around one or more optical fibers coming from a fiber feeder 103. The plastic material to be extruded may be, for example, polybutylene terephthalate (PBT), polypropylene (PP), polyethylene (PE), polycarbonate (PC) or some other plastic polymer. The resulting loose-laid tube may have a cylindrical shape. The extruder can be temperature controlled. In some embodiments, multiple extruders may be involved. In some embodiments, element 105 may correspond to one or more extruders or multiple extruders.

In embodiments where the loose-laid fiber optic cable being manufactured is to be filled with water-resistant (or water-blocking) yarn, recoating line 101 may further comprise one or more yarn feeders for feeding yarn into the loose-laid tube (not shown in FIG. 1). In embodiments where the loose-laid fiber optic cable being manufactured is to be filled with a water-resistant (or water-blocking) gel or electrolyte, recoating line 101 may further comprise a gel or electrolyte injection system for delivering the gel or electrolyte into the loose-laid tubing (not shown). in Fig. 1). Said gel or electrolyte injection system may be incorporated into one or more extruders 105. In some embodiments, recoating line 101 may further comprise a dry (or gel-free) tube diameter control device for controlling the diameter of dry-laid fiber optic cables. tube.

Initially, when the loose-laid tube is extruded by the extruder 105, it has a very high temperature. To efficiently cool the loose-laid tube, the dry-tube loose-laid fiber optic cable 104 may be passed through a flow quench or cooling bath (not shown in FIG. 1). Said flow cooling may extend all the way from extruder 105 to winder 109. Cooling may be controlled by varying the temperature of the cooling water used in flow cooling.

The compression track 106 is used to draw the loose-laid tube extruded by the extruder 105 at a speed that is higher than the predetermined line speed (i.e., faster than the speed of the medium hauler 107 described below) to compensate for the shrinkage of the free-laid tube. stacking fibers as it cools. In other words, the compression track 106 is used to minimize excess fiber length (EFL) and/or tube shrinkage of the loose-laid fiber optic cable being manufactured. Excess fiber length is a measure of how much longer one or more optical fibers are in a loose-laid tube compared to a loose-laid tube containing said one or more optical fibers. The compression track may comprise two moving pressure bands running parallel to each other and positioned opposite each other.

The pressure bands of the compression track 106 may be made of rubber or other material that provides sufficient friction between the pressure bands and the loose-laid extrudable tube. The loose-laid tube can be fed between said pressure bands which compress the tube, and although the friction and high speed of the pressure bands (compared to the speed of the uncompressed line) results in an increase in the speed of the loose-lay tube through the line (i.e. elongation of the tube), this does not affect the transmission rate of one or more optical fibers along this line. The compression caused by the pressure bands must be such that only elastic deformation occurs (and not plastic deformation). In some embodiments, the compression track may comprise, instead of pressure bands, two pressure wheels (made of, for example, rubber) operating in a similar manner to the pressure bands (as described above).

The middle hauler 107 is the next point of physical contact for the loose-laid fiber optic cable 104 after the compression track 106. As a result, the middle hauler 107 also has a significant effect on fiber excess length and tube shrinkage of the loose-laid fiber optic cable being manufactured. tube, and parameters (eg, motor torque) of the middle hauler 107 can be used to control excess fiber length and tube shrinkage.

In addition to the middle hauler 107, the recoating line may further comprise an end hauler 108 to further control the transmission rate of the loose-laid fiber optic cable along the line. The end hauler 108 may be between the middle hauler 107 and the winder 109.

The winder 109 is used to wind the finished fiber optic cable 104 with loose fibers in the tube.

Finally, the recoat line 101 comprises one or more drives 102 for moving and controlling one or more recoat line devices 103, 105-109. One or more actuators 102 may include, for example, one or more electric motors, a hydraulic actuator, and a pneumatic actuator. One or more actuators 102 are powered by and controlled by a control system 116 (or more specifically, an automated programmable logic controller system 119, PLC, control system 116). The operation of one or each of the multiple actuators 102 may be controlled by setting one or more control parameters of the respective actuator. One or more drive control parameters may include, for example, voltage, current, power and/or frequency of a signal applied to the drive. In other words, the operation of the recoat line 101 can be controlled by setting one or more control parameters (in most cases, a plurality of control parameters) of the recoat line, where each control parameter is associated with one or more drives 102 of the recoat line 101. The control parameters can be equal to degrees are called operating parameters.

In addition to or as an alternative to controlling the recoating line 101 by setting one or more control parameters of one or more drives 102, at least one of the other elements 103, 105, 104, 106, 107, 108, 109 of the recoating line 101 can be controlled in according to one or more control parameters. For example, said one or more control parameters of said other items in a recoating line may include one or more temperatures of one or more respective extruders 105, cooling water temperature (in flow cooling), line tension, and fiber feeder tension (i.e., tension on the fiber feeder 103). Line tension can be defined as a metric (value given in Newtons) quantification of how much free-laid tube is stretched during winding. Fiber Feeder Tension can be defined as a metric (value given in Newtons) quantification of how tight the fiber optic cable is at the start of the recoating line (ie, next to the fiber feeder 103).

The control system 116 includes a programmable logic controller (PLC) automated system 119, a process control device 110, a first computing device 111, a user device 112, a machine learning database 113, a history database 114, and a second computing device 115. The first computing device 111 , a machine learning database 113, a history database 114, and a second computing device 115 form a computing system 117 for performing machine learning analysis in accordance with the embodiments. Said control system 116 may be located in the same room as the secondary coating line 101 (ie, it may be a local system). Alternatively, at least some elements (eg, second computing device 115 and databases 113, 114) of control system 116 may be remote elements (eg, remote servers or databases). In some embodiments, some of said remote elements may be cloud-based or other distributed elements. Connections between devices 110-115 of control system 116 shown in FIG. 1 may comprise one or more wired connections (or communication channels) and/or one or more wireless connections (communication channels).

The PLC automation system 119 is used to monitor and control the production of the recoating line 101. More specifically, the PLC automation system 119 can be used to control process parameters of the recoating line 101 and adjust control parameters of the recoating line. In order to provide the above monitoring and control, the PLC automation system 119 can be connected by wired/or wireless communication channels to each individual element 102-109 of the secondary coating line 101 (including flow cooling, which is not shown in Fig. 1) or at least to some of the mentioned individual elements 102–109. In addition, the PLC automation system 119 is connected to the process control device 110. The PLC automation system 119 may be a rugged computing system designed for reliable operation in harsh environments.

With regard to the control of production on the recoating line 101, the measurement data (i.e., process data) issued by each individual element 102-109 of the recoating line 101 connected to the PLC automated system 119, during operation of the recoating line 101, may transmitted periodically or continuously to the automated PLC system 119. The process data collected by the PLC automation system 119 may comprise measured values of one or each of several process parameters of the recoat line 101 (e.g., tension, speed, and/or torque measurements associated with the compression track and/or middle hauler) .

With regard to the control of production on the recoating line 101, the PLC automation system 119 may be configured to at least control the operation of one or more drives 102 of the recoating line 101. In other words, the PLC automation system 119 may provide control signals (or control inputs ) for one or more drives 102 to set their control parameters. One or more process parameters can be tuned by adjusting one or more control parameters of the recoating line 101. For example, increasing the frequency of the AC motor causes an increase in the torque of the AC motor, and therefore also the speed of the moving element that is driven by the AC motor. current (for example, the speed of the compression track).

The process control unit (PSU) 110 of the control system 116 is used to control the parameter set control and alarm systems, and maintain the process parameters received from the PLC automation system 119 in the PSU 110 database (not shown in FIG. 1). A set of parameters in this context can be defined as a set of instructions or steps required to operate a secondary coating line in order to obtain the desired finished product. The PSU 110 may additionally store in said PSU 110 database information on one or more parameter sets and/or nominal settings associated with the secondary coverage line. The PSU 110 may be configured to operate supervisory control and data acquisition (SCADA) software associated with the secondary coverage line 101. In addition, the PSU may be wired/or wirelessly connected to one or more monitors in a control room (not shown). in Fig. 1). Said one or more monitors may be used to display (real time) process information to an operator. The PSU 110 may be configured to transmit or relay (current) process data (i.e., controlled process parameters) received from the PLC automation system 119 to the computer system 117 (more specifically, to the first computer 111 in the illustrated embodiment). implementation) and/or to user device 112. In addition, PSU 110 may be configured to transmit or relay control signals (i.e., messages containing one or more control parameters of the secondary coverage line to be configured) from computer system 117 (more specifically, from the first computing device 111 in the illustrated embodiment) and/or from the user device 112 to the automated PLC system 119.

The first computing device 111 of the computing system 117 is configured to monitor and possibly control the quality of the loose-laid fiber optic cable being manufactured, in accordance with the embodiments. More specifically, the first computing device 111 can estimate the current quality of the loose-laid fiber optic cable using the current process data and the (pre)trained machine learning algorithm stored in the machine learning database 113 that is connected to the first computing device 111. The first computing device 111 may also be electrically connected to the archive database 114, the PSU 110, and the user device 112. The first computing device may be configured to transmit information about the pre-assessed quality of the loose-laid fiber optic cable being manufactured and/or recommendations for possible adjustments to the control parameters of the secondary coating line 101 to improve said pre-evaluated quality on the user's device. The first computing device may also be configured to send control signals to PSU 110 (which may be configured to redirect said control signals to PLC automation system 119.). The first computing device 111 may store in a database (for example, an internal database or a machine learning database 113 or an archive database 114) information about its preliminary quality assessments and/or production line data received from the PSU 110. The first computing the device may also store a buffer of control values in memory and/or have access to a database of PSU 110. PLC automation system 119 and/or PSU 110 may be remotely controlled by computer system 117 (more specifically, first computer device 111 in the illustrated embodiment) and/ or user device 112.

User device 112 may refer to a portable or stationary computing device (equipment, device, terminal). Computing devices that may be used include wireless mobile communication devices operating with or without a Subscriber Identity Module (SIM) in hardware or software, including but not limited to the following types of devices: desktop computer, laptop computer, touch screen computer, mobile phone, smartphone, personal digital assistant (PDA), headset, electronic reader, tablet, game console, laptop, multimedia device, sensor, drive, video camera, car, laptop, telemetry and remote monitoring devices. User device 112 may be connected to computing system 117 via a first wireless or wired communication channel and to PSU 110 via a second wireless or wired communication channel. The user device 112 is configured to at least receive information provided by the first computing device 111 (or in general, the computing system 117) and display said information on the screen (or display) of the user device 112. As mentioned above, the user device 112 can also be used to control the PSU 110 and via the PSU 110 the PLC automation system 119 (i.e. to set the control parameters of the recoating line 101). In order for a user of user device 112 to be able to issue control commands, user device 112 may include at least one user input device (eg, a touch screen, one or more buttons, a keyboard, and/or a mouse). In practice, the user device 112 may be controlled by a master technologist of the recoating line 101 or another person familiar with the operation and management of the recoating line 101.

Finally, the second computing device 115 of the computing system 117 is configured to at least create and train the machine learning algorithm used by the first computing device 111. To achieve this operation, the second computing device is connected to the archive database 114, which contains at least the archived data , containing data on the quality of the fiber optic cable previously manufactured using the recoating line 101 or other similar recoating line, and the corresponding process data for the same recoating line obtained during the manufacture of said fiber optic cable. Said quality data may be based on measurements of off-line quality metrics such as excess fiber length, tube shrinkage, and fiber attenuation. These off-line quality measurements can be performed, for example, in a quality control laboratory, i. e. they are not performed using the elements depicted in FIG. 1 (i.e. using recoating line 101 or control system 116). Element 118 is used to indicate that the quality data based on off-line measurements and stored in the archive database 114 is transferred to the archive database 114 from outside the control system 116 through the external interface of the computer system 117. If the quality data stored in the database archived data 114 correspond to a fiber optic cable manufactured not on the recoating line 101, but on another corresponding recoating line (i.e. another recoating line of the same type and composition), the corresponding process data must also be transferred to the archival database data 114 from outside. The second computing device 115 may be configured to store the trained machine learning algorithm in a machine learning database 113 that both the first and second computing devices 111, 115 have access to.

In some embodiments, the archive database 114 may also contain control parameter data (i.e., control parameters), in other words, control parameters according to which the secondary coating line 101 (or other corresponding secondary coating line) was previously controlled. Said control parameter data may correspond to process data stored in a database (i.e., the control parameters defined in said control parameter data could be used to control the secondary coating line when process parameter values were measured).

Although in Fig. 1 shows an exemplary computing system 117 in accordance with embodiments comprising two separate computing devices 111, 115 and two databases 113, 114, in other embodiments, computing system 117 may comprise a different number of distinct computing devices and/or databases. In some embodiments, computing system 117 may, for example, include a single computing device (performing the functions of first and second computing devices 111, 115) and two databases 113, 114 or one database (storing information from both databases 113, 114) connected to or contained in said single computing device.

On FIG. 2 shows a process according to embodiments for pre-qualification of fiber optic cable during manufacture in a real-time recoating line. The fiber optic cable produced on the secondary coating line can be exactly the fiber optic cable with loose fibers in the tube. The depicted process may be performed by computer system 117 shown in FIG. 1, or any other alternative computing system described with respect to FIG. 1. The process depicted in FIG. 1 may be executed by the first computing device 111 shown in FIG. 1, or any other appropriate computing device (i.e., a computing device having access to at least current process data and a pre-trained machine learning algorithm and connected to a user device) on any other alternative computing system described with respect to FIG. . 1. A computer system or device that implements the process shown in FIG. 1 may be configured to monitor and possibly control (possibly via a PSU and an automated PLC system) a recoating line comprising at least some of the elements described with respect to the recoating line 101 shown in FIG. 1. In some embodiments, it is contemplated that said recoat line may comprise at least a compression track and a middle hauler. Further, the object that performs this process for simplicity will be called a computing system.

Initially, the computing system contains in a first database (e.g., a machine learning database as described in relation to FIG. 1), in block 201, a trained machine learning algorithm for calculating expected values of one or more quality metrics of a fiber optic cable manufactured on recoating line, based on the values of one or more values of the process parameters of the recoating line. In other words, a trained machine learning algorithm takes one or more process parameters as its inputs and outputs expected values for one or more quality metrics as its outputs.

One or more quality metrics (ie, inputs to the trained machine learning algorithm) are metrics that are indicative of at least some aspect of the quality of the fabricated fiber optic cable. The at least one or more performance metrics may correspond to performance metrics estimated in a conventional manner by off-line measurements of the fabricated fiber optic cable. Said one or more quality metrics may include, for example, one or more excess fiber lengths, tube shrinkage and fiber optic cable light attenuation, loose tube dimensions, loose tube dimensional stability, a metric associated with a fiber optic test. compression and metric index associated with the kink test.

In some embodiments, said one or more quality metrics include at least one (preferably all) selected from the group of fiber length excess, tube shrinkage, and fiber optic light attenuation. These three metrics can be considered the most important quality metrics for loose-laid fiber optic cable and are therefore described in more detail here.

Excess fiber length (EFL) can be defined as the length of the fiber in the loose-laid tube divided by the length of the loose-laid tube (sometimes reported as a percentage). A large EFL means that the fiber in the loose-laid tube is coiled. Excessive fiber bending can degrade the performance of the fiber optic cable. Excess fiber length is a metric usually measured off-line for each fiber as a standard quality control procedure. Excess fiber length can also be estimated on line during manufacture by measuring the speed of the fiber at the feeder and the speed of the free-laid tube prior to winding, although these measurements are generally not very accurate.

Tube shrinkage is a phenomenon caused by the relaxation of a plastic free-laid tube after manufacture. The amount of tube shrinkage depends on the process conditions, such as line speed, cooling temperatures and compression track parameters. Tube shrinkage is typically measured off-line for each loose-laid tube manufactured as a standard quality control procedure.

Low light attenuation is the most important characteristic of any fiber optic cable. The attenuation of an optical fiber increases if it is subjected to mechanical stress during manufacture or during winding on a spool. Light attenuation is usually measured out of line as a standard quality control procedure. Light attenuation can be expressed in decibels per meter.

In some embodiments, one or more quality metrics used as inputs to the trained machine learning algorithm include (or consist solely of) an overall quality metric. The overall quality metric is a metric that indicates the overall quality of a manufactured fiber optic cable. An overall quality metric (or its value) may also be referred to as a quality category. The overall quality metric may be defined to have a value selected from a discrete set of numeric values (eg, 0, 1, and 2). Each numeric value of said discrete set may correspond to a specific overall quality rating letter (eg, 0 means "Bad", 1 means "Average", and 2 means "Good"), which can be directly displayed on the screen of the user's device.

An overall quality metric (or category) can be defined as a function of two or more quality metrics whose values can be determined using (off-line) measurements of manufactured loose-laid fiber optic cable (e.g., excess fiber length, tube shrinkage, and/or or attenuation of light in a fiber optic cable). For example, an overall quality metric may be defined as the sum, weighted sum, mean, or weighted average of the two or more quality metrics. The results of said calculation may then be rounded or truncated (eg, to an integer value) to match a value in a predefined discrete set (ie, to match one of the predefined quality categories).

In some embodiments, the computing system may first calculate one or more expected values of one or more performance metrics (e.g., excess fiber length, tube shrinkage, and/or fiber optic attenuation) using a trained machine learning algorithm, and then calculate the expected value an overall quality metric based on said one or more expected values of the one or more quality metrics.

Said one or more process parameters are parameters or properties of various processes of the secondary coating line. Said one or more process parameters may be associated with one or more fiber feeders, extruder, compression caterpillar, middle hauler, final hauler and winder, or any other devices in the secondary coating line. The parameters of the compression track and the middle hauler often have a particularly pronounced effect on the quality of the finished loose-laid fiber optic cable. Thus, one or more process parameters may include one or more process parameters associated with the compression track and/or middle traction device. Said process parameters may include one or more compression track speeds, compression track tension, compression track motor torque and medium traction motor torque, standard deviation of the compression track speed over a predetermined period of time, standard deviation of the compression track tension during a predetermined the predetermined period of time, the standard deviation of the compression track motor torque during the predetermined period of time, and the standard deviation of the medium traction device motor torque during the predetermined period of time.

The creation and training of a trainable machine learning algorithm may be performed by a computing system (more specifically, a second computing device, as described in relation to FIG. 1), for example, as described below in relation to FIG. 3. In other embodiments, the machine learning algorithm to be trained may be created and/or trained by some entity (eg, network node or computing device) other than the computing system performing the processes shown in FIG. 2. As described above, the machine learning algorithm in accordance with the embodiments may be based on, for example, one or more feed-forward neural networks, a Bayesian classifier, a self-organizing map (SOM), or a combination thereof.

The computing system controls in block 202 one or more values of one or more (respectively) process parameters of the secondary coating line during operation of the secondary coating line. Control is necessary because the process parameters of the recoating line may change during the operation of the recoating line due to various factors, such as changes in the environment in which the recoating line operates (for example, an increase in temperature), or wear of line elements Secondary coating (e.g. pressure band of a compression track). The one or more monitored process variables may include at least one or more process variables used as an input to the trained machine learning algorithm. Control may be performed by PSU and PLC automation system as described in relation to FIG. 1. In other words, process data (comprising at least the bottom or several values of one or more process parameters) can be measured and collected during the operation of the recoating line in real time by an automated PLC system and transmitted to the PSU, which redirects the said data to a computing system for analysis (other than placing said data in a PSU database for storage). The values of one or more process parameters may be measured and transmitted to the computer system periodically (using one or more different periods associated with different parameters).

During monitoring, the computing system calculates in block 203 in real time one or more expected values of one or more quality metrics using the trained machine learning algorithm, while the monitored values of one or more process parameters are used as an input parameter of the trained machine learning algorithm.

The computing system provides, at block 204, at least one or more expected values of one or more quality metrics to the user device. In addition, the computing system may also provide monitored (ie, current) values of one or more process parameters corresponding to one or more expected values of one or more quality metrics.

In some embodiments, the output at block 204 may include causing at least one or more expected values of one or more quality metrics to be displayed on the screen of the user's device in real time, assisting the user of that device in managing the secondary coverage line. In some cases (for example, in the case of an overall quality metric), a visual or textual indication of the expected value may be displayed (for example, indicating that the overall quality is “Good” or indicating the value of the quality metric using a visual element, such as a bar chart) instead of or in addition to raw numeric value. Based on the displayed information, the user of the user device can determine whether at least one of the control parameters should be adjusted to cause a change in the process parameters of the secondary coating line, which in turn will lead to an improvement in the quality of the final product. The user device may then transmit, upon appropriate input by the user, control commands to the secondary coating line to set at least one of the control parameters of the secondary coating line.

The term "initiating the display" of information on the screen of the user's device may include in this case and in subsequent embodiments, the implementation of the transmission of commands to the user's device to display the relevant information on the said screen. In other embodiments, the user device may be configured to display any expected quality and/or process information received from the computing system on the screen of the user device automatically (i.e., without being explicitly instructed to do so).

The actions related to blocks 202-204 may be performed by the computer system continuously while the secondary coating line is in operation, as indicated by the arrow connecting block 204 to block 202.

On FIG. 3 shows the process of creating and training the machine learning algorithm described in relation to FIG. 2. This process may be performed by the same computing system, which then performs a pre-evaluation using a (pre-)trained machine learning algorithm in accordance with the embodiments. Considering the system shown in Fig. 1, the processes shown in FIG. 2 may be performed by the first computing device 111 while the processes described with respect to FIG. 3 may be performed by the second computing device 115. Alternatively, the creation and training of the machine learning algorithm may be performed by another entity that is not part of the computing system.

With reference to FIG. 3, the computing system contains in the database (for example, in the archive database) in block 301 archive data containing data on the quality of the optical fiber cable previously manufactured using the recoating line, and the process data of the recoating line obtained during the manufacture of the mentioned fiber optic cable. The quality data comprises a plurality of measured values for one or more quality metrics, and the process data comprises a plurality of measured values for one or more process parameters of a recoating line. One or more quality metrics and one or more process parameters may be defined as described with respect to FIG. 2. Quality data can be collected by standard off-line quality measurements, while process data can be collected by in-line measurements (i.e., by monitoring similar to that described with respect to block 202 in FIG. 2) during operation of the recoating line. .

The computing system initializes, at block 302, a machine learning algorithm to calculate expected values of one or more quality metrics based on the values of one or more process parameters. The machine learning algorithm may be any machine learning algorithm as described above, such as a neural network based algorithm employing one or more feed-forward neural networks and/or one or more recurrent neural networks. Initialization may include defining input parameters (i.e., features) and output parameters (i.e., labels) of the machine learning algorithm and setting or selecting initial values for weights and/or parameters of the machine learning algorithm (e.g., the weights of one or more neural networks ). In this case, the features (i.e., input parameter) of the machine learning algorithm may be defined to match the quality metrics of the manufactured fiber optic cable, and the labels (i.e., output parameter) of the machine learning algorithm may be determined to correspond to the parameters of the manufacturing process. recoating line process. The initial values may be random values, or they may correspond to a predefined set of values resulting in a well-functioning algorithm. Any known initialization technique may be used for initialization at block 302.

The computing system trains, in block 303, a machine learning algorithm using the archived data. In other words, archived data (or part of it) is used as training data for a machine learning algorithm. More precisely, the process data is used as an input to the machine learning algorithm, while the quality data determines the corresponding expected outputs of the machine learning algorithm. Training may include introducing process data (i.e., values of one or more process parameters) into a machine learning algorithm, comparing the expected values of one or more quality metrics produced by the machine learning algorithm with measured values of said one or more quality metrics. in the archived data and adjusting one or more weights and/or parameters of the machine learning algorithm so as to minimize the difference between the expected and measured values. The comparison can be done by evaluating the cost function, which is a measure of how incorrect a machine learning algorithm is in its ability to estimate the relationship between its inputs and outputs. The cost function can be expressed as the difference or distance between the predicted value and the actual value (or the vector of the predicted value and the actual value vector in the case of multiple output variables).

Finally, the computing system, at block 304, stores the trained machine learning algorithm in a machine learning database (or some other database). The computing system (more specifically, the first computing device of the computing system) may then use said trained machine learning algorithm to perform pre-evaluation in accordance with embodiments (eg, as described with respect to FIG. 1).

In some embodiments, the actions related to blocks 301-304 may be performed off-line (ie, when the recoating line is offline or out of service). In other embodiments, the activities related to blocks 301-304 may be performed on the line during normal operation of the secondary coating line.

On FIG. 4 shows an alternative process in accordance with embodiments for pre-qualifying fiber optic cable (loose-laid fibers in the tube) during fabrication in a real-time recoating line. The depicted process may be carried out by the entity described with respect to FIG. 2 as an execution of the process shown in FIG. 2 (for example, a computing system or a certain computing device that is part of it). Typically, the process depicted in FIG. 4 corresponds to a large extent to the process depicted in FIG. 2. Any definitions given in relation to FIG. 2 are applicable or can be combined with the process depicted in FIG. 4 (unless otherwise noted).

With reference to FIG. 4, the initial blocks 401-403 may fully correspond to the blocks 201-203 shown in FIG. 2. Thus, the actions related to the mentioned blocks are not described here for the sake of brevity. Instead, the focus of the description is on the additional/alternative features shown in FIG. 4 as blocks 404, 405.

After the computing system calculates one or more expected values of one or more performance metrics using the trained machine learning algorithm in block 403, the computing system then calculates in block 404 in real time during monitoring one or more optimal values of one or more parameters control of the secondary coating line to improve the quality of the manufactured fiber optic cable. The quality of an optical fiber here may mean in particular the quality defined by one or more expected values of one or more quality metrics (for example, the expected value of a general quality metric or one or more expected values of one or more quality metrics, assessed in the usual way by measurements out of line). The one or more control parameters for which said one or more optimal values are calculated may comprise some or all of the control parameters that can be used (by the PLC automated system) to control the secondary coating line.

Calculation of one or more optimal values for one or more control parameters in block 404 may include, first of all, comparing one or more controlled values of one or more process parameters with one or more corresponding optimal values of one or more process parameters to determine , which process parameters deviate from the optimal values and therefore need to be adjusted. Said optimal values for one or more process parameters may be determined based on historical data (i.e. quality and process data) stored in the archive database or based on a trained machine learning algorithm and/or they may be stored in the database. archival data or machine learning database. With respect to the first option, said archived data may contain data based on off-line measurements and/or data based on previous preliminary quality assessments made by the computing system. In some embodiments, the optimum values for one or more process parameters may simply correspond to the values that have previously resulted in the (highest) quality fiber optic cable. Predefined mappings between process parameters and control parameters may be contained in a database (eg, an archive database, a machine learning database, or a PSU database). Based on these predetermined matches and the optimal values of one or more process parameters in need of tuning, one or more optimal values of one or more control parameters can be calculated (at block 404).

Alternatively, the calculation in block 404 may be performed using a predefined algorithm that takes as an input parameter at least one or more controlled values of process parameters (and, optionally, one or more expected values of one or more quality metrics and/or or controlled control parameters). In other words, the predefined algorithm used at block 404 may correspond to a process model that reflects the internal mechanisms of the manufacturing process on the secondary coating line and models the relationships (or relationships or interactions) between various parameters (i.e., at least between different parameters control and process parameter). The predetermined algorithm may be stored, for example, in a machine learning database or other memory or database connected to or included in the computing system. The predetermined algorithm may be created or built using knowledge of the manufacturing technology and by analyzing data generated by said manufacturing technology (eg, process parameters and/or control parameters). Said predefined algorithm may be generated, for example, based on archived data stored in the archived database, controlled process parameters, controlled control parameters and/or said predefined correspondences (if any). In some embodiments, a trained machine learning algorithm may also or alternatively be used to create the predefined algorithm. Said predefined algorithm may use the trained machine learning algorithm in block 404 calculations. In other embodiments, said predefined algorithm may comprise or correspond to the trained machine learning algorithm. In other words, in some embodiments, a separate predefined algorithm may be provided connecting control parameters not only to process parameters, but also to quality metrics. In these embodiments, the predefined algorithm used in block 404 may be the same as the machine learning algorithm described above.

The computing system initiates, at block 405, tuning during real-time monitoring of one or more control parameters of the secondary coverage line to match one or more optimal values of one or more control parameters (or at least reduce the difference between one or more optimal values and one or more current values of one or more control parameters). Tuning initiation may include sending control data to set one or more control parameters to the PSU, which in turn forwards said control data to a PLC automation system that adjusts the operation of one or more actuators driving the secondary coating line (and possibly the operation other elements of the secondary coating line) respectively. In this particular embodiment, the setting of control parameters can be fully automated, and therefore there is no need for a separate user device to control the recoating line.

The embodiments shown in FIG. 2 and 4 may also be combined such that the computing system provides, between blocks 404, 405, or immediately following block 405, at least one or more expected values of one or more performance metrics to the user device, as described in relation to block 204, shown in FIG. 2. Any additional features described with respect to block 204 shown in FIG. 2 can also be applied in this case.

In some embodiments, the actions related to blocks 404, 405 may be performed only if the computing system determines that at least one or more expected values of one or more performance metrics (calculated in block 403) fall below at least one respective predetermined threshold for the respective at least one quality metric, i.e., if the quality of the manufactured fiber optic cable is found to be sufficiently poor.

In the embodiment depicted in FIG. 4, the setting of control parameters is fully automatic and does not require feedback from the user. However, in some cases, allowing the user to make the final decision on any changes to the control parameters of the secondary coating line can be useful, for example, to improve safety and ensure high product quality. On FIG. 5 shows a signaling scheme in accordance with embodiments for providing said user-defined decision capability. More precisely, in Fig. 5 shows a diagram of the signaling between the computer system, the user device, the PSU, and the PLC automation system, which are the corresponding elements shown in FIG. 1. The computing system shown in FIG. 5 may correspond to any object described with respect to FIG. 2 as executing the process shown in FIG. 2 (for example, a computing system or a specific computing device that is part of it). Typically, the process depicted in FIG. 5 corresponds to a large extent to the processes depicted in FIG. 2 and/or FIG. 4. Any definitions given in relation to FIG. 2 and/or 4 are applicable or can be combined with the process depicted in FIG. 5 (unless otherwise noted).

With reference to FIG. 5, the initial blocks 501-504 may fully correspond to the blocks 401-404 shown in FIG. 4. Thus, the actions related to the mentioned blocks are not described here for the sake of brevity. Instead, the focus of the description is on the additional/alternative features shown in FIG. 5 as elements 505-513.

On FIG. 5 after the computing system calculates in block 504 one or more optimal values for one or more control parameters so as to improve the quality of the manufactured fiber optic cable, the computing system initiates the display on the user device of a request to change one or more current values of one or more control parameters to adjust to one or more optimal values for one or more control parameters. More specifically, initiating a display may include sending control data to the user's device in message 505 for displaying said request on the screen. After receiving the control data in block 506, the user device displays the above request in block 507. A prompt here may be defined as a question or statement that appears on the screen, and which may indicate to the user the options available for selection (in this case, at least the options to change controls and not change controls). Therefore, the operator must approve any control parameter settings.

Upon receipt, at block 508, of a positive user command authorizing the setting(s) of one or more control parameters proposed by the request via the user device's user input device, the user device sends control data to the PSU in message 509 for setting the one or more control parameters. Therefore, after receiving the control data in block 510, the PSU transmits (or redirects) in the message 511 said control data to set one or more control parameters to the PLC automation system. In some embodiments, the PSU may also place control data in the PSU database for storage.

In response to receiving the control data in block 512, the PLC automation system sets one or more control parameters in block 513 in accordance with the control data.

In some alternative embodiments, the computing system may cause the user's device to display, instead of prompting, simply one or more optimal values for one or more control parameters and also additionally current values for one or more control parameters. In such embodiments, the user may not be given the option to quickly change one or more control parameters to match the optimal values, but instead the user may be required to manually set the control parameters using the user's device (using a dedicated recoat line control application).

In some embodiments, the computing system may store in a database (for example, an archive database, a machine learning database, or a PSU database) one or more predefined lists of actions (i.e., actions regarding setting control parameters) that can be performed when a deviation of a certain process parameter is detected. Based on said one or more predefined lists of actions, the computing system may initiate displaying one or more options for setting one or more control parameters to improve the quality of the fiber optic cable on the user's device. For example, if the engine torque standard deviation value is found to be much higher than the optimum value, then the computing system may cause the user device to display information about one or more actions that the operator can take to reduce the instability of the process indicated by high standard deviation.

After the training of the machine learning algorithm is completed (for example, as described in relation to Fig. 3) and the trained machine learning algorithm has been used for a certain time to evaluate the quality of the manufactured fiber optic cable, it may be useful to retrain the machine learning algorithm using the latest data on quality. On FIG. 6A and 6B show two different retraining processes in accordance with embodiments. The processes depicted (or only one of them) may be performed by any of the entities described with respect to FIG. 2 as the execution of the process shown in FIG. 2, for example, by a computer system for a preliminary assessment of the quality of an optical fiber cable during manufacture on a secondary coating line, or by a certain computer device that is part of this system. The computing system, or the particular computing device included within it, may be configured to perform one or both of the processes depicted in FIG. 6A and 6B. For example, the process shown in Fig. 6A may be performed intermittently using the first period, the process shown in FIG. 6B may be performed periodically using a second period (preferably much longer than the first period), or only upon receipt of a request, for example, from the user's device.

With reference to FIG. 6A, the computing system receives, at block 601, one or more sets of archive data via an external interface of the computing system. Each set of archived data may be similar to the archived data described in relation to the above embodiments. In other words, each archived data set contains data on the quality of a loose-laid fiber optic cable previously manufactured using one of the secondary coating lines (i.e., a secondary coating line manufacturing fiber optic cable, the quality of which must be pre-evaluated by a computer system) , and another corresponding recoating line (for example, a special recoating reference line) and process data of the same recoating line obtained during the manufacture of said loose-tube fiber optic cable. Here (similar to the above embodiments), the quality data contains a plurality of measured values for one or more quality metrics, and the process data contains a plurality of measured values for one or more process parameters of the recoating line (or equivalently, mentioned other relevant secondary coating lines). If said one or more archive data sets are composed of a plurality of archive data sets, said plurality of archive data sets may be received at block 601 as a single transmission or as a plurality of transmissions received periodically or from time to time. With respect to the latter option, any new archive data can be committed to the computing system as soon as it becomes available, but receiving a single set does not necessarily trigger any retraining process. Instead, the relearning process shown in FIG. 6A may be triggered based on a timer (eg, after a predetermined time or period after the initial training or last update) and/or based on the amount of (new) historical data collected by the computing system.

Upon receipt of one or each of the multiple archive data sets at block 601, the computing system, at block 602, stores the received archive data set in a database of the computing system. In some embodiments, block 602 may be omitted.

The computing system generates, at block 603, an updated trained machine learning algorithm by retraining the trained machine learning algorithm (stored in the machine learning database) using the one or more historical data sets. In other words, in this embodiment, the currently trained machine learning algorithm is used as a starting point for retraining. In this way, the current trained machine learning algorithm can be further fine-tuned to improve the accuracy of the preliminary estimates. Retraining in block 603 may include, for example, if the machine learning algorithm is based on one or more neural networks, adjusting one or more values of one or more weight coefficients of said one or more neural networks.

After retraining is completed and the updated trained machine learning algorithm is generated in block 603, the computing device stores the updated trained machine learning algorithm in block 604 in a machine learning database. Further, the computing system may use said updated trained machine learning algorithm for preliminary quality assessment (eg, in accordance with any of the above embodiments). In the depicted embodiment, it is assumed that updating the trained machine learning algorithm is repeated immediately upon receipt of the new archive data set(s) (as described above with respect to block 601) and thus the process proceeds from block 604 back to block 601.

With reference to FIG. 6B, the actions related to blocks 611, 612, 614 may be exactly the same as those described above with respect to blocks 601, 602, 604 shown in FIG. 6A and are therefore not repeated here for the sake of brevity. In other words, the difference between the two depicted processes lies solely (or predominantly) in the training step depicted in blocks 603, 613. If block 603 shown in FIG. 6A relates to fine-tuning the currently trained machine learning algorithm, the computing system generates, at block 613, an updated trained machine learning algorithm by initializing a new machine learning algorithm to compute expected values of one or more performance metrics based on the values of one or more process parameters, and trains this new machine learning algorithm using at least one or more archived datasets (possibly also using other archived data stored in the archived database). In other words, learning is actually restarted from scratch. Thus, the creation of the updated trained machine learning algorithm in block 613 may proceed in a similar manner as described for the original creation of the trained machine learning algorithm, i.e., as described in relation to block 303 shown in FIG. 3 (although, of course, using a different set or sets of archived data as an input parameter). This type of machine learning algorithm update may allow more significant changes to the machine learning algorithm compared to the process at block 603 shown in FIG. 6A. For example, the process shown in Fig. 6B may cause the updated trained machine learning algorithm to have a different topology (e.g., a different topology for one or more neural networks) compared to the previous trained machine learning algorithm, as opposed to, for example, causing changes only in the weights one or more neural networks. Thus, if it turns out that the preliminary estimates produced by the trained machine learning algorithm are no longer reliable, even despite periodic retraining attempts in accordance with FIG. 6A, the process shown in FIG. 6B may be performed to generate a restructured trained machine learning algorithm that is more suitable for pre-qualification of the fiber optic cable currently being manufactured on the secondary coating line. In the embodiment depicted in FIG. 6B, unlike the embodiment shown in FIG. 6A, it is assumed that the creation of a new trained machine learning algorithm is not automatically repeated (ie, this process terminates at block 614).

The retraining performed by the computing system in accordance with FIG. 6A and/or 6B may be carried out in manual, semi-automatic or fully automatic mode. In most cases, the computer system can perform semi-automated retraining, which means that the historical data (containing the relevant quality and process data) obtained from off-line measurements is collected continuously, but the trained machine learning model is retrained according to with Fig. 6A and/or 6B periodically or as needed. Initiating relearning on demand may include, for example, initiating relearning upon receipt of a request by the user device, where the user device may transmit said request upon receipt of a predetermined user command via the user device's user input device. Alternatively, the computing system may be programmed to be fully automatic such that the trained machine learning algorithm is updated, for example, once a week using all the new archive data available (i.e., all data collected since the last update). The nature of fiber optic cable production using a secondary coating line is such that spool changes and line shutdowns occur several times a day. These stops or breaks in the production process can be used to download an updated machine learning model for use (and possibly also to create an updated trained machine learning algorithm).

Training and/or retraining of a computing system in accordance with embodiments may be supervised or unsupervised. In supervised learning, a historical dataset (containing quality and process data) is used to train a machine learning algorithm to produce a desired calculation result using certain process data as an input parameter. With unsupervised learning, the computing system itself tries to distribute the data into a certain number of feature groups. Feature groups may or may not correspond to the above quality categories, which may be known to the user. Quality data can be used to test how well unsupervised learning can differentiate feature groups, but it is not used extensively in machine learning. For example, multilayer perceptrons can be trained in a supervised way, while self-organizing maps can be trained in an unsupervised way.

On FIG. 7 shows an apparatus 701 configured to perform the functions described above in connection with a computing system, such as computing device 117 shown in FIG. 1, or in relation to a computing device, such as the first computing device 111 of the computing system 117 in FIG. 1. This device may be an electronic device containing electronic circuits. This device may be a single (network) entity or a plurality of individual entities. This device may include a control circuit 720 such that at least one processor and at least one memory 730, including computer program code (software) 731, configured with at least one memory and computer program code (software), with at least one processor, provided that the device executes any of the above-described embodiments of the computing system.

Memory 730 may include at least one database 732. Said at least one database 732 may comprise at least a machine learning database and/or an archive database as described in relation to the above embodiments. In other embodiments, the one or more machine learning databases and the archive database, as described in relation to the embodiments, may be external databases or database servers that the device 701 has access to through communication interfaces 710. The memory 730 may also contain other databases (eg, a database for storing at least controlled process parameters), which may or may not be associated with the described quality pre-evaluation functionality in accordance with the embodiments.

With reference to FIG. 7, the control circuit 720 may include a machine learning circuit 721 configured to provide device functionality to pre-assess the quality (or more specifically, one or more quality metrics) of a manufactured (loose-laid) fiber optic cable and additionally to communicate the results. preliminary quality assessment on the user's device in accordance with any of the presented embodiments. In some cases, the machine learning circuit 721 may also be configured to trigger the adjustment of one or more secondary coverage line control parameters. For example, machine learning circuit 721 may be configured to perform at least some of the processes shown in FIG. 1-4, 6A and 6B and/or some of the processes performed by the computing system in FIG. 5. In some other embodiments, the control circuit 720 may be divided into two or more separate circuits.

The device 701 may also include interfaces (communications) 710 containing hardware and/or software for implementing the ability to establish a data connection in accordance with one or more data transfer protocols. The communication interface may provide the device with communication capabilities for communicating with, for example, one or more user devices, a secondary coating line process control device, and/or one or more external databases or database servers. One or more communication interfaces 710 may provide communication capabilities to device 701 for exchanging data in a cellular communication system and establishing communication with one or more network nodes and one or more terminal devices.

The communication interface may 710 may comprise standard well known components such as an amplifier, filter, frequency converter, (de)modulator, encoder/decoder circuitry, and one or more antennas.

The memory 730 of the device 701 described with respect to FIG. 7 may be implemented using any suitable storage technology, such as semiconductor memories, flash memory, magnetic storage devices and systems, optical storage devices and systems, ROM and removable storage.

As used herein, the term "circuit" may refer to one or more of the following: (a) implementations of hardware circuitry only (e.g., implementations in analog and/or digital circuitry only) and (b) a combination of hardware circuits and software, for example (if applicable ): (i) a combination of analog and/or digital hardware circuitry with software/firmware and (ii) any parts of the hardware processor(s) with software (including digital signal processor(s)), software, and memory (i) that work together to cause a device such as a mobile phone or a server to perform various functions) and (c) hardware circuit(s) and/or processor(s), such as a microprocessor(s) or part of a microprocessor( s) that needs software (such as firmware) to operate, but may not have software if it is not required to operate.

This definition of schema applies to all uses of the term in this application, including any claims. As a further example, used in this application, the term "circuit" also covers the implementation of simply a hardware circuit or a processor (or several processors), or a part of a hardware circuit or a processor and its (or their) accompanying software and/or firmware. The term "circuitry" also encompasses, for example, and when applicable to a particular claim element, a baseband or processor integrated circuit for a mobile device or a similar integrated circuit in a server, a cellular network device, or other computing or network device.

In an embodiment, at least some of the processes described in connection with FIG. 2-5, 6A and 6B may be carried out by a device containing appropriate means for performing at least some of the described processes. Some exemplary means for executing processes may include at least one of the following: detector, processor (including dual-core processors and processors with multiple cores), microprocessor, digital signal processor (DSP), controller, microcontroller, receiver, transmitter, encoder, decoder , memory, random access memory (RAM), read only memory (ROM), software, firmware, display, user interface, display circuit, user interface circuit, user interface software, display software, circuit, antenna, antenna circuitry, application specific integrated circuit (ASIC), digital signal processing device (DSPD), programmable logic device (PLD), and field programmable gate array (FPGA). With respect to firmware or software, implementations in accordance with embodiments may be implemented using modules of at least one chipset (procedures, functions, etc.) that perform the functions described here. The software codes may be stored in a storage device and executed by processors. The storage device may be implemented within the processor or external to the processor. In the latter case, it may be communicatively coupled to the processor by various means known in the art. In an embodiment, the at least one processor, memory, and computer program code form processing means or comprise one or more computer program code portions for performing one or more operations in accordance with any of the embodiments shown in FIG. 2–5, 6A and 6B, or operations thereof.

Embodiments as described may also be performed as computational processes defined by a computer program or portions thereof. Embodiments of the methods described in connection with FIG. 2-5, 6A and 6B may be accomplished by executing at least one portion of a computer program containing appropriate instructions. A computer program may be in source code, object code, or some intermediate form, and it may be stored on a medium of some sort, which may be any object or device capable of containing the program. For example, the computer program may be stored on a distributable computer program medium that is readable by a computer or processor. A computer program medium can be, for example, but not limited to, a recording medium, a computer memory, a read only memory, an electrical carrier signal, a telecommunications signal, and a software distribution package. The medium for the computer programs may be a non-volatile medium. The software coding for executing the embodiments as shown and described is within the comprehension of one skilled in the art.

While the invention has been described above with reference to an example in accordance with the accompanying drawings, it is to be understood that the invention is not limited thereto and may be modified in several ways within the scope of the appended claims. Thus, all words and expressions should be interpreted broadly and are intended to be illustrative and not restrictive of an embodiment. One skilled in the art will appreciate that as technology advances, the concept of the invention may be implemented in a variety of ways. In addition, one skilled in the art will appreciate that the described embodiments may, but need not, be combined with other embodiments in various ways.

One skilled in the art will appreciate that as technology advances, the concept of the invention may be implemented in a variety of ways. The invention and its embodiments are not limited to the examples described above, but may be varied within the scope of the claims.

Claims (47)

1. A quality control method for a loose-laid fiber optic cable during manufacturing on a secondary coating line, this method includes the following steps: storing in the machine learning database a trained machine learning algorithm for calculating expected values of one or more quality metrics of a loose-laid fiber optic cable manufactured on the recoating line based on the values of one or more process parameters of the recoating line; one or more values of one or more process parameters of the secondary coating line are controlled by means of a computer system during operation of the secondary coating line; one or more expected values of one or more quality metrics are calculated by the computing system during real-time monitoring using the trained machine learning algorithm, while the controlled values of one or more process parameters are used as an input parameter of the trained machine learning algorithm; And at least one or more expected values of one or more quality metrics are output by the computing system to the user's device. 2. The method of claim 1, wherein at least one or more expected values of one or more quality metrics are output to a user device, comprising: initiating, with the help of a computing system, the display of one or more expected values of one or more quality metrics on the screen of the user's device in real time, helping the user of this device in managing the secondary coating line. 3. The method of claim 2, wherein at least one or more expected values of one or more quality metrics are output to the user device, further comprising: using the computing system, they initiate displaying one or more values of one or more process parameters of the secondary coating line on the screen of the user's device in real time, helping the user of this device in managing the secondary coating line. 4. The method according to any one of the preceding claims, wherein the recoating line operates in accordance with one or more control parameters of the recoating line, the method further comprising: calculating by the computer system during real-time monitoring one or more optimal values of one or more control parameters to improve the quality of the manufactured fiber optic cable with loose fibers in the tube, as determined by one or more expected values of one or more quality metrics; And initiating with the help of a computer system in real time in the course of monitoring the setting of one or more control parameters to bring them into line with one or more optimal values of one or more control parameters. 5. The method according to any one of paragraphs. 1-3, in which the recoating line operates in accordance with one or more control parameters of the recoating line, this method further includes: calculating by the computer system during real-time monitoring one or more optimal values of one or more control parameters to improve the quality of the manufactured fiber optic cable with loose fibers in the tube, as determined by one or more expected values of one or more quality metrics; And initiating, with the help of the computing system, display on the screen of the user's device in response to the calculation of one or more optimal values of one or more control parameters, a request to change one or more current values of one or more control parameters to bring them into line with one or more optimal values of one or more control parameters. 6. The method of quality control of fiber optic cable with loose fibers in the tube during manufacture on the secondary coating line, this method includes the following steps: storing in the machine learning database a trained machine learning algorithm for calculating expected values of one or more quality metrics of a loose-laid fiber optic cable manufactured on the recoating line based on the values of one or more process parameters of the recoating line; one or more values of one or more process parameters of the secondary coating line are controlled by means of a computer system during operation of the secondary coating line; one or more expected values of one or more quality metrics are calculated by the computing system during real-time monitoring using the trained machine learning algorithm, while the controlled values of one or more process parameters are used as an input parameter of the trained machine learning algorithm; calculating, during real-time monitoring, one or more optimum values of one or more control parameters of the secondary coating line to improve the quality of the manufactured fiber optic cable with loose fibers in the tube, as determined by one or more expected values of one or more quality metrics; And initiating, in the course of real-time monitoring, adjustment of one or more control parameters of the secondary coating line to bring it into line with one or more optimal values of one or more control parameters. 7. The method according to any one of paragraphs. 4-6, in which one or more control parameters of the recoating line contains one or more control parameters of one or more drives that drive the recoating line, and / or one or more of the temperatures of one or more respective extruders, the temperature of the cooling water, line tension and fiber feeder tension on the secondary coating line, said one or more control parameters of one or more drives contain for one or each of several drives at least one voltage, current, power and frequency of the control signal applied to the corresponding drive. 8. A method according to any one of the preceding claims, further comprising generating a trained machine learning algorithm by performing the following steps: store in the archive database, the archive data containing the quality data of the fiber optic cable with loose fibers in the tube, previously manufactured using one of the recoating lines and the other recoating line, and the process data of the same recoating line obtained during manufacturing of said fiber optic cable with loose fibers in the tube, wherein the quality data contains a plurality of measured values of one or each of several quality metrics, and the process data comprises a plurality of measured values of one or each of several process parameters of said one of the secondary lines coating and other secondary coating line; initializing, by means of the computing system, a machine learning algorithm for calculating expected values of one or more quality metrics based on the values of one or more process parameters; teaching, by means of the computing system, a machine learning algorithm using the archived data; And storing, by means of the computing system, the trained machine learning algorithm in a machine learning database. 9. The method of claim 8, further comprising: one or more sets of historical data are received by the computer system through the external interface of the computer system, each set of archive data contains data on the quality of the fiber optic cable with free laying of fibers in the tube, previously manufactured using one of the secondary coating lines and the other secondary coating line , and process data of the same recoating line obtained during the manufacture of said fiber optic cable with loose fibers in the tube, wherein the quality data contains a plurality of measured values of one or each of several quality metrics, and the process data contains a plurality of measured values of one or each of several process parameters of said one of the recoating lines and the other recoating line; using the computing system, an updated trained machine learning algorithm is created by performing the following actions: retraining the trained machine learning algorithm using one or more historical datasets, using the trained machine learning algorithm as the starting point for retraining, or initializing a new machine learning algorithm to calculate expected values of one or more quality metrics based on the values of one or more process parameters and training a new machine learning algorithm using at least one or more historical data sets; And storing, by the computing system, the updated trained machine learning algorithm in a machine learning database. 10. The method according to claim 8 or 9, in which the control and all calculations during the control are performed by the first computing device of the computing system, and the creation of the trained machine learning algorithm is carried out by the second computing device of the computing system. 11. The method of any one of the preceding claims, wherein the trained machine learning algorithm is based on one or more feed-forward neural networks, one or more recurrent neural networks, a Bayesian classifier, a self-organizing map, or a combination thereof. 12. The method of claim 11, wherein the trained machine learning algorithm is based on at least one or more feed-forward neural networks comprising one or more layered perceptron networks and/or one or more convolutional neural networks, and/or on one or more recurrent neural networks containing one or more networks with long-term and short-term memory and/or one or more recurrent convolutional neural networks. 13. The method of any one of the preceding claims, wherein the one or more quality metrics include one or more of fiber length excess, tube shrinkage, and fiber optic light attenuation. 14. The method of any one of the preceding claims, wherein the one or more quality metrics include a general quality metric or a general quality category indicative of the overall quality of the loose-laid fiber optic cable being manufactured. 15. The method of any one of the preceding claims, wherein the recoating line comprises a compression track, and one or more process parameters include one or more of the following: compression track speed, compression track tension, compression track motor torque, compression track speed standard deviation tracks for a predetermined period of time, the standard deviation of the compression track tension for a predetermined period of time, and the standard deviation of the compression track motor torque for a predetermined period of time. 16. The method of claim. 15, in which the secondary coating line contains the average traction device located after the compression caterpillar, and one or more process parameters include the torque of the engine of the average traction device and / or the standard deviation of the torque of the engine of the average traction device for a predetermined period of time. 17. A quality control device for a loose-laid fiber optic cable during manufacture in a recoating line, comprising means for carrying out the method of any one of the preceding claims. 18. A non-volatile computer-readable medium that stores instructions that, when executed by a computing device, cause the computing device to execute the method according to any one of paragraphs. 1-16. 19. The secondary coating line control system, comprising: a computing system configured to perform the method according to any one of paragraphs. 1-16; a user device comprising at least a screen and a user input device connected to the computing system by a communication channel; an automated programmable logic controller (PLC) system configured to at least control process parameters of the secondary coating line and adjust control parameters of the secondary coating line; And a process control unit (PSU) connected by communication channels between the computer system and the automated PLC system and between the user device and the automated PLC system, while the PSU is configured to relay the controlled process parameters from the automated PLC system to the computer system and the user device, as well as with the possibility of relaying control signals from the computer system and user device to the automated PLC system.

Images