PLC-based optical fiber winding tension control method and system

Abstract

The invention discloses a PLC-based optical fiber winding tension control method and system, and the method comprises the steps: obtaining the remaining length of an optical fiber according to a set and real-time winding length through a length decreasing model, switching a winding unit to an ending preparation state and triggering a generator to decelerate when the remaining length is smaller than a threshold value, and obtaining the position information of a winding mechanism through a photoelectric sensor after the ending speed is reached; a motor speed correction coefficient is obtained through a position loop speed correction algorithm according to the angle difference and the target angle difference, and a control instruction is generated. And if the angle reaches or exceeds the target angle, starting the neural network prediction model to obtain a shutdown advance, and outputting a shutdown signal by the PLC according to the shutdown advance and a preset path model to stop the motor with low vibration. The PLC is utilized to integrate various sensing and intelligent algorithms, high-precision tension control in the full-winding process is achieved, the problems that in the traditional technology, tension control is not stable, and problems are prone to occurring at the end are effectively solved, the optical fiber winding quality and efficiency are improved, and the method adapts to various types of optical fibers and technological requirements.

Description

Technical Field

[0001] The invention belongs to the technical field of optical fiber winding, and in particular relates to a method and system for controlling the tension of optical fiber winding based on PLC. Background Art

[0002] In the field of optical fiber manufacturing, optical fiber winding is a key production link. Traditional optical fiber winding equipment usually relies on simple mechanical structures and relatively basic control methods to achieve the winding process. For example, in terms of tension control, mechanical tensioners or simple electrical control circuits are often used. Mechanical tensioners have limited adjustment accuracy and poor adaptability to optical fibers of different wire diameters and materials. Its tension adjustment mainly relies on the friction or elastic deformation of mechanical components, and it is difficult to maintain stable and accurate tension output throughout the winding process. Once the wire diameter, material or winding speed of the optical fiber changes, it is difficult for the mechanical tensioner to make accurate adjustments in time, which easily leads to tension fluctuations during the optical fiber winding process.

[0003] Simple electrical control circuits are also unable to cope with complex winding conditions. They often lack the ability to fully monitor and effectively integrate and analyze various dynamic parameters (such as the remaining length of the optical fiber, the change in motor speed, the position of the winding mechanism, etc.) during the winding process. This means that when facing some winding tasks with special requirements, such as the winding of high-precision optical fiber products with extremely high tension stability requirements, or complex processes that require frequent speed and tension changes during the winding process, traditional electrical control circuits cannot meet the precise tension control requirements, which in turn affects the quality and performance consistency of optical fiber winding products.

[0004] In addition, at the end of fiber winding, traditional control methods have many drawbacks. Usually, the winding motor is simply stopped according to a fixed deceleration program or relying on the operator's experience, lacking precise control and prediction of the winding mechanism position, fiber tension, and the dynamic characteristics of the entire system. This can easily cause a sudden change in fiber tension at the end, resulting in fiber breakage or uneven winding, which seriously reduces the product yield and production efficiency.

[0005] Based on the above analysis, the main technical problem solved by the present invention is that the existing optical fiber winding technology has defects such as low precision, poor adaptability and easy problems in the finishing stage in terms of tension control, which makes it difficult to meet the demand of the modern optical fiber manufacturing industry for high-quality products. Summary of the invention

[0006] The technical problem to be solved by the present invention is to provide a method and system for controlling the tension of optical fiber winding based on PLC, which realizes high-precision tension control from the beginning to the end of the optical fiber winding process through accurate length calculation model, advanced motor deceleration control algorithm, precise position detection and speed correction strategy and predictive shutdown control mechanism based on neural network, effectively reduces tension fluctuation, avoids optical fiber breakage, improves product yield and quality consistency, and can adapt to different types of optical fibers and various winding process requirements, thereby improving the intelligence level and production efficiency of the entire optical fiber winding production process.

[0007] In order to solve the above technical problems, the technical solution adopted by the present invention is: A method and system for controlling the tension of optical fiber winding based on PLC, the steps are as follows: S1: According to the preset total length of the optical fiber and the real-time winding length information fed back by the current encoder, the remaining length of the optical fiber is calculated by the length reduction model; S2: If the remaining length of the optical fiber is less than or equal to the preset winding trigger length threshold, the winding control unit switches from the normal winding state to the winding preparation state, and obtains a winding control unit signal that the winding control unit is about to perform the winding operation; S3: triggering the motor deceleration control module according to the finishing preparation state signal, using a preset deceleration curve control algorithm to smoothly reduce the motor speed, and determining that the motor running speed has reached the finishing winding speed value; S4: When the motor runs at a low speed and the length of the wound optical fiber reaches the value obtained by subtracting a safety margin from the final trigger length, the winding control unit obtains the current position information of the winding mechanism through a photoelectric sensor installed at a preset reference position, and determines the angle of the optical fiber relative to the preset reference position in the final winding stage; S5: According to the difference between the current angle of the real-time winding mechanism and the target angle, the winding control unit calculates the winding motor speed correction coefficient through the speed correction algorithm based on the position loop, obtains the control instruction for the fiber tail winding, and performs precise position control on the tail winding process; S6: If the output of the position control calculation model shows that the current angle of the winding mechanism has reached or exceeded the target angle, the winding control unit starts the preset neural network prediction model, and the neural network prediction model analyzes and learns from the historical operation data to obtain the advance amount information of the terminal parking position, and determines the amount of stop signal that needs to be issued in advance in the terminal control link; S7: Based on the stop advance of the finishing control link given by the prediction model and the preset finishing path model, the PLC outputs a stop signal to the driver, causing the winding motor to stop in a low-vibration deceleration mode to avoid fiber breakage and obtain a complete winding product.

[0008] Preferably, the sub-steps of S1 are: S2.1 If the optical fiber surplus length data value is less than or equal to the preset tail trigger threshold length value; S2.2 The winding control unit status data is updated to 1; S2.3 According to the winding control unit state data, when the winding control unit state data is 1, the winding control unit switches to the finishing preparation state; S2.4 The winding control unit obtains the finishing preparation status data; S2.5 The winding control unit enters the finishing preparation state; S2.6 The winding control unit generates the pre-trigger signal data for the closing operation; S2.7 The winding control unit pre-trigger signal data according to the closing operation; S2.8 The winding control unit runs the self-test module of the equipment to obtain the normal signal data of the equipment self-test; S2.9 Use the normal signal data of the equipment self-test as the precondition for the final closing operation; S2.10 The tailing signal is triggered and output, and the winding equipment performs the optical fiber tailing operation.

[0009] Preferably, the sub-steps of S2 are: S2.1 If the optical fiber surplus length data value is less than or equal to the preset tail trigger threshold length value; S2.2 The winding control unit status data is updated to 1; S2.3 According to the winding control unit state data, when the winding control unit state data is 1, the winding control unit switches to the finishing preparation state; S2.4 The winding control unit obtains the finishing preparation status data.

[0010] Preferably, the sub-steps of S3 are: S3.1 generates a high level output according to the signal trigger of the end preparation state. After the motor controller receives the high level, it obtains the information and starts the control; S3.2 The starter motor deceleration control module processes the end preparation state signal and transmits it to the programmable logic controller through the bus to obtain the instruction data set; S3.3 Obtain the function relationship expression y=f(x) of the preset deceleration curve by table lookup method, where the independent variable x represents time and the dependent variable y represents the motor speed, and obtain the expected speed set of the motor; S3.4 obtains the real-time motor speed information measured by the encoder and transmits it to the data processor through the serial port to obtain the actual running speed value; S3.5 The Kalman filter algorithm is used to reduce the noise of the actual running speed value of the motor, and the actual running speed value after noise reduction is connected to the PID controller as the feedback input quantity. The PID parameters are adjusted iteratively to obtain the optimal PID control instruction; S3.6 The PID controller combines the motor speed value obtained by the preset deceleration curve with the actual operating speed value after noise reduction. If the difference between the two is greater than the set value, the alarm module is triggered to issue an early warning. S3.7 uses BP neural network algorithm to establish a prediction model. The prediction model is trained using the actual operating speed set obtained through PID control. It determines whether the predicted speed value meets the final winding speed value and outputs a stop signal.

[0011] Preferably, the sub-steps of S4 are: S4.1 uses an encoder installed on the motor to periodically record the speed and generate a data stream in combination with time parameters; S4.2 This data flows through the data buffer pool for temporary storage; S4.3 The neural network prediction model adjusts its own parameters in real time through a self-learning process based on historical operating rate data and establishes an operating rate determination benchmark; S4.4 The model compares the current data with the judgment benchmark obtained through learning, determines the operation rate value and defines different working stages for different values. The operation rate less than a certain value is defined as the low-speed operation stage; S4.5 The running status value enters the judgment reference data stream as a status tag for subsequent analysis and processing; S4.6 Calculate the number of winding turns according to the preset parameters of the optical fiber, set the initial turn value, combine the information collected by the photoelectric sensor to obtain the final turn value and the circumference of the winding reel, multiply them to obtain the total length at the current moment, and then subtract them to obtain the actual remaining length; S4.7 The end trigger length setting value is subtracted from the safety margin setting value to calculate the comparison length; S4.8 uses the preset rules to determine the actual remaining length and the comparison length value, and obtains the result of whether the remaining length is within the end trigger range; S4.9 If the length falls within the range, a tail trigger instruction is generated; S4.10 The preset rule obtains the closing trigger instruction. The closing stage control logic parses the data packet in the instruction to obtain the control parameters and performs data classification processing, and distributes the classified control instructions to the designated device; S4.11 When the photoelectric sensor detects and obtains trigger information, the trigger information is stored in the buffer area; S4.12 A photoelectric sensor installed at a preset reference position of the winding device detects the periodic rotation signal of the winding reel and collects the rotation data sequence; S4.13 The photoelectric sensor converts the position data into an analog electrical signal corresponding to one rotation period of the winding device; S4.14 Perform frequency domain analysis on the acquired analog electrical signal using Fourier transform to obtain harmonic components and main frequency; S4.15 Calculate the total length of the optical fiber and the number of turns of the device, calculate the two to obtain the corresponding length information of each turn, and divide the remaining length of the optical fiber by the length of each turn to obtain the quotient and remainder; S4.16 The remainder value is multiplied by the duration corresponding to each circumference and the phase corresponding to each circumference to obtain the converted angle value; S4.17 Obtain the total phase angle and convert the remainder into a radian value based on the specific reference point; S4.18 A numerical calculation model is pre-set, and the numerical calculation model collects the numerical value after Fourier transformation and clock data in real time as input data, and uses differential calculation to obtain the numerical change over a period of time; S4.19 Time series acquisition All data obtained from S4.1-S4.18 above are fused; S4.20 The time correlation is obtained by comparing the photoelectric sensor information with the reference time series through the time window, and then the phase offset value is obtained by combining the mechanical structure motion amplitude information corresponding to the reference time series, so as to obtain the time parameter; S4.21 Calculate the product of the time value and the angular velocity value to obtain the total angle; S4.22 According to the time parameter, determine the current instantaneous angular velocity of the motor and the corresponding optical fiber length value; S4.23 The system updates the trigger length threshold of the closing phase in real time according to the fiber length, and uses the fiber type data in the established database as the judgment object; S4.24 obtains the final parameter value of each material, and sends the final parameter value to the motion controller through the data interface; S4.25 A comprehensive comparative analysis is conducted in combination with the mechanical structure characteristic values ​​of the optical fiber winding equipment; S4.26 determines the actual position deviation value through the output value of the photoelectric sensor and the internal setting control parameters; S4.27 The recurrent neural network predicts the winding position information for the next period and obtains multiple future time prediction points; S4.28 The predicted point is subtracted from the current value to obtain the position change trend and change speed, and the trend data is used as the basis for feedback control data in real time; S4.29 adjusts the motor working parameters through the fuzzy control algorithm according to the deviation value, and the photoelectric sensor monitors a preset time value to synchronize the time and value data and store them in the historical database; S4.30 The historical database then classifies and stores data of different storage types, thereby achieving speed adjustment.

[0012] Preferably, the sub-steps of S5 are: S5.1 obtain the current angle information of the winding mechanism and set the target angle information according to the real-time detection; S5.2 Calculate the angle difference based on the current angle information and the target angle information. When the angle difference exceeds the preset range, enter the early warning database; S5.3 reads the angle difference from the warning database. If the angle difference is lower than the preset lower limit, the speed correction step is skipped; S5.4 If the angle difference is higher than the preset upper limit range, an alarm message corresponding to the angle difference is generated and the display module is notified to display the message, the control instruction generation is paused, and at the same time, the current angle and the target angle are read from the warning database; S5.5 calculating the angle difference by a control algorithm based on a position loop to obtain a speed correction coefficient; S5.6 Correct the current output torque of the winding motor according to the speed correction coefficient to obtain corrected real-time torque control information; S5.7 adaptively compensate the real-time torque control information according to the preset winding target length of the optical fiber to obtain a final control instruction; S5.8 Establish an environment perception database for the real-time winding environment, and send the final control command to the winding control actuator. The actuator receives the final control command to control the winding motor to perform the optical fiber finishing winding work; S5.9 If the final state parameter in the final control command shows stop, the actuator is suspended from winding, and the control command update is suspended, then the working data update information in the environmental perception database is obtained, the new winding target length information is obtained, and the corresponding control command information is read from S5.5, and the command is resent; S5.10 If the final status parameter in the final control instruction does not indicate stop, continue with this step; S5.11 performs online detection according to the real-time data information base of the winding mechanism execution status, determines the motion status of the final winding work and saves it.

[0013] Preferably, the sub-steps of S6 are: S6.1 obtain the current angle information of the winding mechanism from the winding mechanism, and determine the difference between the current angle information and the target angle information; S6.2 If the difference between the current angle information and the target angle information is greater than the set value, the winding control unit outputs a signal to the neural network prediction model; S6.3 The neural network prediction model obtains the signal of the winding control unit, and then retrieves the historical operation data from the database for analysis; S6.4 According to the analysis results, the advance information of the end parking position is obtained; S6.5 Determine the amount of advance of the stop signal issued by the terminal control according to the advance information of the terminal parking position; S6.6 The winding control unit receives the stop signal and controls the winding mechanism to perform corresponding operations; S6.7 The database receives and stores angle information, signal quantity and winding mechanism information.

[0014] Preferably, the sub-steps of S7 are: S7.1 Obtain the stop advance data of the final stage output by the prediction model, and obtain the motor control command by fusing the stop advance data and the preset final stage path model information; S7.2 The winding machine driver obtains the fused motor control instruction. According to the current winding motor speed data, the driver controls the winding motor to run to the target speed, which is lower than half of the initial winding speed. S7.3 By presetting the timing, the timing parameters come from the database. The optical fiber status sensor group obtains the optical fiber diameter data in operation, and the optical fiber tension sensor group obtains the real-time tension data. The operation status is estimated using the historical time data. When the optical fiber diameter change amplitude is greater than the set threshold at at least one moment, or when the optical fiber tension is greater than the threshold at at least one moment, the optical fiber status monitoring module makes a shutdown judgment and triggers the optical fiber abnormal status data marking action. The abnormal mark is identified by the neural network to determine the specific abnormal mark, and then the optical fiber abnormal status data marking result is recorded; S7.4 At the closing stage, collect the optical fiber abnormal status data marking results obtained in step S7.3, and analyze whether the number of all abnormal markings is 0; S7.5 During the finishing stage, a convolutional neural network modeled based on the relationship between the fiber finishing process parameters is used to determine whether secondary winding is required; If there are abnormal optical fiber status data marking results and product defects, the secondary winding control link is carried out. According to the defect type and location data, combined with the defect elimination strategy, a new winding control instruction set is planned, and the actual motor operation control instruction is calculated based on the control instruction set and the product defect location coordinate value. Otherwise, the secondary winding operation is not performed; S7.6 The winding machine driver obtains the control signal after secondary winding; S7.7 transmits the signal to the motor drive unit of the winding machine through the signal transmission device using industrial Ethernet; S7.8 The drive unit determines the winding mode and low-vibration mode of the winding machine in the finishing stage and performs winding processing; S7.9 The winding machine executes this control signal in the finishing stage until the product integrity sensor determines that the product winding is complete, and calculates the integral value through the integrity calculation function. The function is I = ∫t0 to t1 f(x)dx, where I represents the integrity, f(x) is the function of integrity with respect to time, and t0 and t1 represent the time of the winding stage; If the winding result is incomplete, a control message is generated and reaches the motor execution unit through the transmitter, and the winding continues; S7.10 After obtaining the completely wound product, combined with the prediction model, the actual winding product finishing data is obtained; S7.11 Conduct data statistics on the process of finishing winding to obtain the actual result of the finishing data and generate a data comparison chart; S7.12 Data storage and analysis: Store the data in the finishing stage and use it for subsequent analysis and improvement to ensure that each winding process can learn from historical data, continuously optimize the control algorithm, and improve the product quality and production efficiency.

[0015] Preferably, the sub-steps of step S7.4 are: S7.4.1 If the number of anomaly marks is greater than 0, based on the anomaly marks, re-plan the parameters of the finishing control link; S7.4.2 According to the shutdown advance amount of the finishing control link obtained in S7.4.1, generate a control signal and send it to the motor control link, and the motor control link adjusts the control curve; S7.4.3 If the number of anomaly marks is 0, do not trigger the mechanism for regenerating the control link parameters, and keep the control link parameters unchanged as the result of S7.4.1, and continue with the subsequent operations.

[0016] A fiber optic winding tension control system based on PLC adopts the described fiber optic winding tension control method based on PLC.

[0017] The present invention can achieve the following beneficial effects: The present invention constructs a length decreasing model to calculate the remaining length of the optical fiber in real time, and combines the preset finishing trigger length threshold to control the winding operation to smoothly transition from normal winding to the finishing preparation state. The present invention adopts a deceleration curve control algorithm to reduce the motor speed, obtains the position information of the winding mechanism through an optoelectronic sensor, calculates the difference between the current angle and the target angle, and applies a speed correction algorithm based on the position loop to accurately control the finishing process. It also introduces a neural network prediction model to learn the shutdown position advance amount from historical data, combines the preset finishing path model, and accurately outputs a shutdown signal to achieve low-vibration deceleration stop, ensuring the smooth end of the optical fiber winding process and the integrity of the product. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The present invention will be further described below in conjunction with the accompanying drawings and embodiments: Figure 1 It is a control logic diagram of a method for controlling the tension of optical fiber winding based on PLC; Figure 2 A schematic diagram of the signal transmission process of a PLC-based optical fiber winding tension control method. DETAILED DESCRIPTION

[0019] The preferred solution is Figure 1 to Figure 2 As shown, a method and system for controlling the tension of optical fiber winding based on PLC, Embodiment 1: A method for controlling the tension of optical fiber winding based on PLC, comprising the following steps: Step S1, according to the preset total length of the optical fiber and the real-time winding length information fed back by the current encoder, the remaining length of the optical fiber is calculated by using the established length reduction model.

[0020] Step S2: If the remaining length of the optical fiber is less than or equal to the preset termination trigger length threshold, the winding control unit switches from the normal winding state to the termination preparation state, and obtains a signal that the winding control unit is about to perform a termination operation.

[0021] The sensor continuously monitors the data value of the remaining optical fiber length; obtains the data value of the remaining optical fiber length and the preset tail trigger threshold length value. Compares the data value of the remaining optical fiber length and the preset tail trigger threshold length value; determines if the data value of the remaining optical fiber length is less than or equal to the preset tail trigger threshold length value. If the data value of the remaining optical fiber length is less than or equal to the preset tail trigger threshold length value; the winding control unit state data is updated to 1.

[0022] According to the winding control unit state data, when the winding control unit state data is 1, the winding control unit switches to the finishing preparation state; and the finishing preparation state data is obtained. The winding control unit is in the finishing preparation state; and the finishing operation pre-trigger signal data is generated. According to the finishing operation pre-trigger signal data; the equipment self-test module runs a self-test and obtains the equipment self-test normal signal data. The equipment self-test normal signal data is used as the final finishing operation precondition; the finishing signal is triggered and output, and the winding equipment performs the optical fiber finishing operation.

[0023] The sensor continuously monitors the residual length data value of the optical fiber at a frequency of 10 times per second, and transmits the data to the data processing center in real time. The data processing center pre-sets a tail trigger threshold length value, such as 10 meters. The data processing center uses a sliding average algorithm to average the received 5 consecutive data points to reduce the measurement error. The average length at the current moment is the average value of the optical fiber length in the past 5 seconds (including the current moment), and a smoother optical fiber residual length data value is obtained. Assume that the residual length obtained by the current average calculation is 15 meters. Through programming, the data processing center compares the smoothed optical fiber residual length data value 15 meters obtained by the average calculation in real time with the preset tail trigger threshold length value 10 meters. Through simple numerical comparison logic, it is judged and compared with the latest monitoring data 15 meters. After confirming that the optical fiber residual length data value 15 meters is greater than the preset tail trigger threshold length value 10 meters, the winding control unit state data remains 0. Through numerical comparison, it is found that if the average calculated smoothed fiber surplus length data value is 8 meters, and 8 meters is less than the preset closing trigger threshold length value of 10 meters, the winding control unit state data is immediately updated to 1, and the state switching mechanism is triggered at the same time. The system transmits the data to the winding control unit. The winding control unit controls the motor to reduce the speed based on the updated state data. Specifically, the original motor speed is 1000 revolutions per minute, and now it gradually decreases to 50 revolutions per minute, entering the closing preparation state. When the fiber surplus approaches the set value, the control system can use a linear or exponential function to trigger and generate a pulse signal with a frequency of 1 Hz by comparing the latest data with the functional relationship between the threshold value. The signal lasts for 5 seconds and is then processed by the system as the closing operation pre-trigger signal data.

[0024] The pre-trigger signal data of the closing operation activates the equipment self-check module, which checks the status of all sensors through the preset diagnostic program and sends a read request to each sensor. All sensors respond within 1 second and feedback their own status. If they are all normal, the equipment is operating normally. After the system receives the normal signal data of the equipment self-check, the analysis module can parse this normal signal data, verify whether the data feature code after parsing meets the expectations, and judge that it meets the expectations to confirm that all equipment is normal. After finally confirming that all feedback data of the equipment self-check module are in normal state, a pulse signal with a frequency of 5 Hz and a duty cycle of 50% is sent to the winding equipment as a closing signal. After receiving this signal, the winding equipment performs the preset closing operation, such as winding the remaining optical fiber neatly to the predetermined position to complete the closing work of the entire optical fiber.

[0025] Step S3, triggering the motor deceleration control module according to the finishing preparation state signal, using a preset deceleration curve control algorithm to smoothly reduce the motor speed, and determining that the motor running speed has reached the finishing winding speed value.

[0026] According to the trigger of the closing preparation state signal, a high level output is generated. After the motor controller receives the high level, it obtains information to control the start. The start motor deceleration control module processes the closing preparation state signal and transmits it to the programmable logic controller through the bus to obtain the instruction data set. Through the table lookup method, the functional relationship expression y=f(x) of the preset deceleration curve is obtained. The independent variable x represents time and the dependent variable y represents the motor speed. The expected speed set of the motor is obtained. The encoder measures the motor speed information in real time and transmits it to the data processor through the serial port to obtain the actual running speed value. The Kalman filter algorithm is used to reduce the noise of the actual running speed value of the motor, and the actual running speed value after noise reduction is connected to the PID controller as the feedback input. The PID parameters are adjusted iteratively to obtain the optimal PID control instruction. The PID controller integrates the motor speed value obtained by the preset deceleration curve and the actual running speed value after noise reduction. If the difference between the two is greater than the set value, the alarm module is triggered for early warning. The BP neural network algorithm is used to establish a prediction model. The model is trained using the actual running speed set obtained by PID control, and the predicted speed value is judged to meet the closing winding speed value, and a stop signal is output.

[0027] According to the end-ready state signal trigger, a high-level output is generated, for example, the voltage is higher than 3 volts. After receiving the high-level signal, the motor controller interprets the signal as a start command, executes the corresponding control logic, and drives the motor to start decelerating. The start motor deceleration control module processes the end-ready state signal, for example, transmits the signal to the programmable logic controller (PLC) in a specific message format through the CAN bus protocol. After receiving the message, the PLC interprets the message content and extracts key information, such as the current operation stage identifier and the start deceleration flag, and stores it in the specified data register, and finally summarizes it into a command data set.

[0028] The contents of the instruction data set are as follows: including data type code, version number, length, data item, etc. Taking the data in this table as an example, the ending speed is 8, the distance is 432, the running time is 17 seconds, and the initial speed is 10. According to the above parameters, the motor start deceleration is set to 0x111 of data item 1, and the starting speed is set to 0x110 of data item 2, and the version number is defined as 0x1. Then, through the preset deceleration curve table, the function expression that meets the starting state is found as follows: y=-5x^2+10, x represents time (in seconds), y represents motor speed (in revolutions per minute), The expected speed set of the motor is calculated as {5, 8, 5, 2, 0}, which correspond to the situations when x={1, 2, 3, 4, 5} respectively. The calculation at this time is based on the total running time of 17 seconds. This speed set represents the speed values ​​corresponding to 5 different time points in the process from the beginning of deceleration to the complete stop of the motor under ideal conditions. At the same time, the motor speed is collected every 10 milliseconds through the encoder to obtain a series of speed data such as {4, 1, 6, 2, 1} and transmitted to the processor through the serial port. The Kalman filter algorithm is used, and the initial estimate value is set to 5, the prediction error covariance is 1, and the measurement noise covariance is 1. The Kalman filter formula is used: prior estimate x(k|k-1)=x(k-1|k-1)+0, state prediction mean square error P(k|k-1)=P(k-1|k-1)+1, and the Kalman gain K(k)= P(k|k-1) / (P(k|k-1)+ 1) is calculated. The motor speed estimate x(k|k) = x(k|k-1)+ K(k)*(z(k)- x(k|k-1)) and the estimated mean square error P(k|k) = (1-K(k))*P(k|k-1) are further calculated to filter the measured speed. The filtered speed is connected to the PID controller as feedback, and the proportional coefficient is set to 8, the integral coefficient to 2, the differential coefficient to 1, and the maximum deviation threshold to 5. The adjusted PID parameters are obtained by iterative calculation, which are 0, 15, and 05 respectively. When the PID control algorithm integrates the motor speed of 5 and the actual speed of 4 after filtering, and the difference is 1, which is less than the set threshold of 5, there is no need to trigger an alarm. However, when the difference reaches or exceeds 5, the alarm module is activated to output an alarm.

[0029] Using the actual speed set obtained after adjustment as a training sample, a BP neural network model is initialized, for example, it contains 3 input nodes (time, target speed, actual speed), 5 hidden layer nodes, and 1 output node (predicted speed). The Sigmoid activation function is used for 500 training iterations. The predicted speed value is judged. If it has reached and stabilized at the rated speed requirement of 5 rpm in the final stage, a stop command is output to stop the motor. The current status data of the equipment is reported to the monitoring system through the network protocol to ensure that the operating status is grasped in real time by the remote monitoring platform, so as to facilitate centralized status supervision and problem diagnosis, provide real-time and reliable information support for the formulation of production scheduling plans and equipment maintenance plans, and further optimize the scheduling of production processes.

[0030] Step S4, when the motor runs at a low speed and the length of the wound optical fiber reaches the final trigger length minus a safety margin, the winding control unit obtains the current position information of the winding mechanism through a photoelectric sensor installed at a specific position, and determines the angle of the optical fiber relative to the specific reference position in the final winding stage.

[0031] According to the above business content and related attributes, a detailed business solution process is proposed. The following are the specific steps: The encoder installed on the motor periodically records the speed and generates a data stream in combination with the time parameter. This data stream is temporarily stored in the data buffer pool. The neural network model adjusts its own parameters in real time through the self-learning process according to the historical running speed data and establishes the running speed judgment benchmark. The model compares the current data with the judgment benchmark obtained by learning, determines the running speed value, and defines different working stages for different values. The running speed is less than a certain value and is defined as a low-speed running stage.

[0032] The running status value is used as a status tag to enter the judgment benchmark data stream for subsequent analysis and processing. The number of winding turns is calculated according to the preset parameters of the optical fiber, and the initial turn value is set. The closing turn value and the circumference of the winding reel are obtained by combining the information collected by the photoelectric sensor. The total length at the current moment is multiplied, and then the actual remaining length is subtracted. The closing trigger length setting value is subtracted from the safety margin setting value to calculate the comparison length. The preset rules are used to determine the value of the actual remaining length and the comparison length to obtain the judgment result of whether the remaining length is within the closing trigger range.

[0033] If the length falls within the range, a tail trigger instruction is generated. The tail trigger instruction is obtained according to the preset rules. The tail control logic parses the data packet in the instruction to obtain the control parameters and performs data classification processing, and distributes the classified control instructions to the specified device. When the photoelectric sensor monitors and obtains the trigger information, the data is stored in the buffer area. The photoelectric sensor installed at a specific position of the winding device detects the periodic rotation signal of the winding disk and collects the rotation data sequence. The photoelectric sensor converts the position data into an analog electrical signal of the period of one rotation of the winding device.

[0034] The analog electrical signal is analyzed in the frequency domain using Fourier transform to obtain the harmonic components and main frequency.

[0035] Calculate the total length of the optical fiber and the number of turns of the device, calculate the two to obtain the corresponding length information of each turn, and divide the remaining length of the optical fiber by the length value of each turn to obtain the quotient and remainder. Multiply the remainder value by the corresponding duration of each circumference and the corresponding phase of each circumference to obtain the converted angle value. Obtain the total phase angle, and convert the remainder part into a radian value based on a specific reference point. Preset a numerical calculation model, which collects the values ​​after Fourier transformation and clock data as input data in real time, and uses differential calculation to obtain the numerical changes over a period of time.

[0036] Time series acquisition The above multiple steps obtain data and perform data fusion. The time correlation is obtained by comparing the photoelectric sensor information with the reference time series through the time window, and then the phase offset value is obtained by combining the mechanical structure motion amplitude information corresponding to the reference time series, thereby obtaining the time parameter. The product of the time value and the angular velocity value is calculated to obtain the total angle. According to the time parameter, the instantaneous angular velocity of the current motor and the corresponding fiber length value are determined. The system updates the trigger length threshold of the closing stage in real time according to the fiber length, and uses the fiber type data in the established database as the judgment object. The closing parameter value of each material is obtained, and the value is sent to the motion controller through the data interface. A comprehensive comparative analysis is performed in combination with the mechanical structure characteristic value of the optical fiber winding equipment. The actual position deviation value is determined by the output value of the photoelectric sensor and the internal setting control parameter. The recurrent neural network predicts the winding position information of the next period and obtains multiple future time prediction points.

[0037] The predicted point is subtracted from the current value to obtain the position change trend and change speed, and the trend data is used as the basis for feedback control data in real time. The motor working parameters are adjusted through the fuzzy control algorithm according to the deviation value, and the photoelectric sensor monitors a preset time value to store the time and value data synchronously in the historical database. The historical database then classifies and stores data of different storage types. This achieves speed adjustment.

[0038] The encoder installed on the motor records the speed at a frequency of 1000 times per second. For example, at a certain moment, the recorded speed is 3000 rpm, and the data stream is generated in combination with the time parameter 10:00:00 at this time. This data stream is temporarily stored in a data buffer pool with a capacity of 10MB. The neural network model records the data of the past month as training samples based on the historical running rate data in the past month, for example, once per minute, and records a total of 43,200 data. Assuming that there are a total of 40,000 valid data, these data are imported into the self-built neural network model. After 1000 rounds of cyclic training, a total of 1000 rounds of cyclic training are carried out. During the training process, the parameters are adjusted in real time by using the method of cross entropy loss function as the loss function, and a running rate judgment benchmark is established. For example, one of the reference values ​​in the benchmark is 2800 rpm. The model compares the current data 3000 rpm with the judgment benchmark obtained by learning, and determines that the running rate value is a normal value. The neural network will define different working stages for different values. The running speed is less than a certain value, such as less than 1000 rpm, which is defined as the low-speed running stage. The running status value enters the judgment benchmark data stream as a status label for subsequent statistics, regression or fault diagnosis.

[0039] The total length of the optical fiber is set to 5000 meters. According to the preset parameters of the optical fiber, the length of the optical fiber required for each winding is set to 2 meters, so the number of winding turns is calculated to be 2500, and the initial turn value is set to 0. Combined with the mark information of each turn collected by the photoelectric sensor, the ending turn value is 2400 turns. For example, the circumference of the winding disk is 2 meters. The number of ending turns is multiplied by the circumference of the winding disk. After winding 2400 turns, the total length of the wound optical fiber is 4800 meters. 5000 minus 4800 can calculate the actual remaining length of 200 meters. Subtract the ending trigger length setting value, such as 250 meters, from the safety margin setting value, such as 20 meters, to obtain a comparison length value of 230 meters. Then use the comparator to compare 200 with 230. Since 200 meters is less than 230 meters, the judgment result that the remaining length is in the ending trigger range will be obtained, and a ending trigger instruction will be generated.

[0040] According to the generated closing trigger instruction, the control logic parses the data packet in the instruction to obtain the control parameters, such as length, number of turns, speed, etc., and performs data classification processing to distribute the control instruction to the execution device, such as the motor, PLC controller, etc. When the photoelectric sensor monitors the trigger information generated by the winding disk rotating to the preset position, the data is stored in a buffer area with a capacity of 5MB. The photoelectric sensor installed at a specific position of the winding device detects the periodic rotation signal of the winding disk, and generates a pulse signal for each turn, and collects the rotation data sequence at a frequency of 100 per second. The photoelectric sensor converts the specific position coordinate data when the winding rotates one circle, such as the change of the x coordinate is 10 cm, or the change of the y coordinate is 15 cm, etc., according to the detected position coordinate data, into an analog electrical signal corresponding to the rotation period of the winding device, and then uses a 12-bit analog-to-digital converter to convert the signal. The obtained analog electrical signal is analyzed in the frequency domain using Fourier transform to obtain harmonic components and main frequency. For example, the fast Fourier transform algorithm is used for analysis to identify the main frequency, harmonic amplitude and phase and other data. The total length of the optical fiber is 5000 meters, divided by the total number of winding turns of 2500, and the length of the optical fiber corresponding to each turn is calculated to be 2 meters, and the quotient obtained by dividing the remaining length of the optical fiber of 200 meters by the value of the optical fiber length of 2 meters per turn is 100, and the remainder is 0. The remainder 0 value is multiplied by the corresponding time length of each turn of 2 meters and the corresponding phase of each turn of circumference, such as π, and the converted angle value is 0. Convert the remainder of the total phase angle into a radian value based on a specific reference point, such as π. A numerical calculation model is set in advance. This numerical calculation model collects the numerical value after fast Fourier transform and clock data in real time as input data, uses differential calculation to obtain the numerical change within a period of time, and then uses methods such as Runge-Kutta to perform numerical calculations within a fixed time window, and updates the parameters of the Fourier transform to obtain numerical values ​​such as phase angle and frequency.

[0041] Time series acquisition The above multiple steps obtain data and merge these data into a unified data stream. Then, through a time window of 10 units, compare the photoelectric sensor information with the reference time series to obtain the value of time correlation. For example, the correlation between the two sequences is 95.

[0042] Combined with the reference time series corresponding to the mechanical structure movement amplitude information, for example, the maximum movement amplitude is 50 mm, the phase offset value is obtained. The product of the time value and the angular velocity value is calculated to obtain the total angle. Then the time information read by the encoder is 1 second, and the corresponding angular velocity is 100 revolutions per second. They are multiplied to obtain the total rotation angle of the winding turntable is 100 turns, and then the radian value corresponding to each turn is multiplied by this value to obtain the total rotation angle value.

[0043] According to the time parameter, for example, the time is 2 pm, the instantaneous angular velocity of the current motor is obtained, such as 120 revolutions per second and the corresponding fiber length value of 3 meters. According to the current fiber length of 200 meters, the system updates the trigger length threshold of the closing stage, and uses the fiber type data in the established database, such as its elastic coefficient, density, etc., for numerical calculation as the judgment object, and combines the mechanical structure characteristic values ​​of the fiber winding equipment, such as the diameter of the reel, the maximum speed and other information for comprehensive comparison and analysis. The closing parameter value of each material is obtained, and this value is sent to the motion controller through the data interface of the hub and the optical fiber. The motion controller combines the real-time position data, such as the current 100th circle, and compares the output value of the photoelectric sensor with the internal setting control parameter to determine the actual position deviation value, such as the deviation between the current position and the theoretical position is 5 cm. The recurrent neural network is used to predict the winding position information of the next period, and the Archimedean spiral is used to establish a model of the future time point. The recurrent neural network generates the data of the predicted points of the next 10 moments based on the previous 1000 historical data. For example, the predicted coordinates of the next five moments are (1,2), (2,3), (3,4), (4,5), and (5,6). Subtract the predicted points from the current values ​​using vector calculation methods to obtain the position change trend and speed, and use the trend data as the basis for feedback control data in real time.

[0044] According to the 5 cm deviation value obtained in the previous step, the corresponding numerical processing is performed through the fuzzy control algorithm to adjust the motor working parameters, such as obtaining the proportional, integral, and differential coefficients of the speed adjustment, which are 5, 3, and 2 respectively. The photoelectric sensor monitors a preset time value, and the time and value data are synchronously stored in a historical database with a capacity of 1GB. The historical database classifies and stores data of different storage types, such as time data stored in the time database and value data stored in the value database. This data is used to guide speed adjustment.

[0045] Step S5, according to the difference between the current angle of the real-time winding mechanism and the target angle, the winding control unit calculates the winding motor speed correction coefficient through the speed correction algorithm based on the position loop, obtains the control instruction for the optical fiber tail winding, and performs precise position control on the tailing process.

[0046] According to real-time detection, the current angle information of the winding mechanism and the set target angle information are obtained. According to the current angle information and the target angle information, the angle difference is calculated. When the angle difference exceeds the preset range, the early warning database is entered. The angle difference is read from the early warning database. If the angle difference is lower than the preset lower limit range, the speed correction step is skipped. If the angle difference is higher than the preset upper limit range, an alarm message corresponding to the angle difference is generated and the display module is notified to display the information. The control instruction generation is paused. At the same time, the current angle and the target angle are read from the early warning database. The angle difference is calculated by a control algorithm based on the position loop to obtain a speed correction coefficient.

[0047] The current winding motor output torque is corrected according to the speed correction coefficient to obtain corrected real-time torque control information. The real-time torque control information is adaptively compensated according to the preset winding target length of the optical fiber to obtain a final control instruction.

[0048] An environmental perception database is established for the real-time winding environment, and the final control instruction is sent to the winding control actuator. The actuator receives the final control instruction to control the winding motor to perform the fiber tail winding work. If the tail state parameter in the final control instruction shows stop, the actuator pauses the winding, and the control instruction update is paused, then the working data update information in the environmental perception database is obtained, the new winding target length information is obtained, and the corresponding control instruction information is read from the fifth step, and the instruction is resent.

[0049] If the final state parameter in the final control instruction does not indicate stop, continue this step. Perform online detection based on the real-time data information library of the winding mechanism execution state, determine the motion state of the final winding work and save it.

[0050] Real-time detection of the winding mechanism angle, for example, the current angle is 30 degrees, and the target angle is set to 90 degrees. The angle difference is calculated to be 60 degrees. Assuming that the preset angle difference range is plus or minus 5 degrees, the current angle difference of 60 degrees far exceeds the preset range, so the 60-degree difference data is stored in the early warning database. Next, the system reads this difference from the early warning database. 60 degrees far exceeds the preset upper limit of 5 degrees, so a specific alarm message is generated, such as: "Warning: The angle deviation is too large, the actual deviation is 60 degrees, which has exceeded the preset upper limit of 5 degrees!" And send this information to the display module, so that the staff can see the specific warning content.

[0051] At the same time, pause the generation of new control instructions, and obtain the current angle of 30 degrees and the target angle of 90 degrees from the early warning database again for subsequent processing. The 60-degree angle difference is calculated through the proportional integral differential control algorithm based on the position loop. For example, set the proportional coefficient Kp to 5, the integral coefficient Ki to 1, and the differential coefficient Kd to 01. According to the control algorithm formula: output = Kp * error + Ki * integral error + Kd * differential error, a speed correction coefficient is calculated, such as 6. According to the speed correction coefficient 6, the current winding motor output torque is corrected. Assuming that the current output torque is 10 Nm, the corrected torque is 10 * 6 = 6 Nm, thereby obtaining the corrected real-time torque control information of 6 Nm.

[0052] Combined with the preset target length of the optical fiber winding, for example, the target length is 100 meters left, adaptive compensation is performed to reduce the error. This can be achieved by adding an adaptive component, using an adaptive least mean square algorithm (LMS) to adjust the speed coefficient according to the deviation from the target winding length.

[0053] For example, given the current length and the target length, we can define the length deviation as follows: Length deviation = current length - target length, the current length is 110, then, length deviation = 110 - 100 = 1 Then we will use this length error as the input of the adaptive component. For example, the real-time torque control information obtained previously is multiplied by an adaptive compensation coefficient of 1, and the final control instruction is 6 Nm * 1 = 6 Nm. Then an environmental perception database is established to record environmental data such as temperature 25 degrees Celsius, humidity 60%, air pressure 101 kPa, as well as control instruction sending time, equipment number and other information. The final control instruction of 6 Nm is sent to the actuator. After the actuator receives the torque instruction of 6 Nm, it accurately controls the winding motor to work.

[0054] If the final state parameter in the final control command shows stop, the actuator stops the winding motor, the control command update is suspended, and the new winding target length is updated to 50 meters from the environmental perception database. Therefore, adaptive compensation is performed from the previously corrected real-time torque control information of 6 Nm to reduce the error. Assuming that the adaptive compensation coefficient is 2, the final control command of 2 Nm is obtained.

[0055] Send the updated data as the final control instruction to the actuator, and let it perform the task according to the new parameters. Otherwise, continue the current step. For example, if the ending state parameter is 1, then maintain a torque output of 6 Nm. At the same time, use the information recorded in the real-time data information library of the winding mechanism execution status, such as motor speed, winding length, winding tension 2 Newtons, etc., to analyze these data to determine whether the current winding state is normal, and save the results of each analysis, such as "Device A is in normal winding working state at 10:30 on October 27, 2023, with a cumulative winding length of 500 meters."

[0056] Step S6, if the output of the position control calculation model shows that the current angle of the winding mechanism has reached or exceeded the target angle, the winding control unit starts the preset neural network prediction model. The neural network model analyzes and learns from the historical operation data to obtain the advance information of the terminal parking position, and determines the amount of stop signal that needs to be issued in advance in the terminal control link.

[0057] The current angle information of the winding mechanism is obtained from the winding mechanism, and the difference between the current angle information and the target angle information is determined. If the difference between the current angle information and the target angle information is greater than the set value, the winding control unit outputs a signal to the neural network prediction model. The neural network prediction model obtains the signal of the winding control unit, and retrieves the historical operation data from the database for analysis. According to the analysis results, the advance amount information of the end parking position is obtained. According to the advance amount information of the end parking position, the amount of the stop signal issued in advance by the end control is determined. The winding control unit receives the stop signal and controls the winding mechanism to perform the corresponding operation. The database receives and stores the angle information, signal amount and winding mechanism information.

[0058] The winding mechanism detects its current angle in real time through the built-in angle sensor, for example, the current angle is 80 degrees. The preset target angle of the winding control unit is 180 degrees. The comparator in the control unit calculates these two values ​​and obtains an angle difference of 100 degrees. The system sets a threshold, such as 10 degrees, and only when the angle difference is greater than this threshold will the next operation be triggered. In this example, 100 degrees is greater than 10 degrees, so the winding control unit outputs a high-level signal, such as 5 volts, to the neural network prediction model. After the neural network prediction model receives this 5-volt signal, the activation model starts working. It will query the database and retrieve the historical data of the past 1,000 winding operations. These data include the starting angle, target angle, actual parking angle of each winding, and the advance amount of the stop signal corresponding to each operation, etc. Then, they can be filtered and analyzed by timestamp to exclude data that occurs at abnormal times. The data includes the date and time of the record. For example, if the current time is 12:00 noon on January 1, 2024, the data from 8:00 to 12:00 noon on that day will be filtered out, for example, 30 data, and time decay will be added to each data through the timestamp. The last one near noon will be included in the data of the winding machine in the latest time period at a ratio of 80%. The second to last time before the last time will be attenuated at a ratio of 40%, and the third to last will be attenuated at a ratio of 20%. The fourth to last will be attenuated at 10%.

[0059] All the remaining data are eliminated and then calculated through the neural network model. This model can be established through the long short-term memory network model to obtain the calculated results.

[0060] This neural network prediction model can analyze the parking position of the winding mechanism under similar conditions for the obtained data, and then analyze the advance amount of the stop signal to obtain the required prediction value. In this example, it may be found that the stop signal needs to be issued 5 degrees before the target angle is reached in order to stop accurately at the 180-degree position. Based on this result, the stop advance amount is set to 5 degrees. Once the winding control unit detects that the current angle reaches 175 degrees (180 degrees minus 5 degrees advance amount), it immediately sends a stop signal to the winding mechanism. After receiving the signal, the winding mechanism stops accurately when it reaches the target position of 180 degrees through the built-in braking system, and controls the winding machine to gradually decelerate until it finally stops accurately at the target angle of 180 degrees, ensuring that the winding angle meets the process requirements. This series of operation data, including the starting angle of 80 degrees, the target angle of 180 degrees, the actual parking angle of 180 degrees, the stop advance amount of 5 degrees, and the motor current of the winding mechanism, the winding speed and other data information, are all recorded and stored in the database in real time to provide data support for subsequent prediction and optimization.

[0061] Step S7, based on the stop advance of the finishing control link given by the prediction model and the preset finishing path model, the PLC outputs a stop signal to the driver, so that the winding motor stops in a low-vibration deceleration mode to avoid causing optical fiber breakage and obtain a complete winding product.

[0062] According to the prediction model output of the final phase stop advance data, the following steps are performed: the prediction model output of the final phase stop advance data is obtained, and the motor control instruction is obtained by fusing the stop advance data and the preset final phase path model information.

[0063] The winding machine driver obtains the fused motor control instructions. According to the current winding motor speed data, the driver controls the winding motor to run to the target speed, which is lower than half of the initial winding speed. Through the preset timing, the timing parameters come from the database. The fiber state sensor group obtains the fiber diameter data in operation. At the same time, the fiber tension sensor group obtains the real-time tension data. The operation status is estimated using the historical moment data. When the fiber diameter change amplitude is greater than the set threshold at at least one moment, or when the fiber tension is greater than the threshold at at least one moment, the fiber state monitoring module makes a shutdown judgment and triggers the fiber abnormal state data marking action. The abnormal mark is identified by the neural network to determine the specific abnormal mark, and then the fiber abnormal state data marking result is recorded. In the closing stage, the fiber abnormal state data marking result is collected to analyze whether the number of all abnormal marks is 0. If the number of abnormal marks is greater than 0, the operation parameter planning of the closing control link is re-performed based on the abnormal mark.

[0064] According to the advance amount of shutdown in the finishing control link, a control signal is generated and sent to the motor control link, and the motor control link adjusts the control curve. If the number of abnormal marks is 0, the control link parameter regeneration mechanism is not triggered, the control link parameters remain unchanged, and the subsequent operations continue. In the finishing stage, a convolutional neural network obtained by modeling the relationship between the fiber finishing process parameters is used to determine whether secondary winding is required. If there is a fiber abnormal state data marking result and there is a product defect, the secondary winding control link is carried out. According to the defect type and position data, combined with the defect elimination strategy, a new winding control instruction set is planned, and the actual motor operation control instruction is calculated and generated according to the control instruction set and the product defect position coordinate value. Otherwise, the secondary winding operation is not performed. The winding machine driver obtains the control signal after the secondary winding. Through the signal transmission device, it is transmitted to the winding machine motor drive unit using industrial Ethernet. The drive unit determines the winding mode and low vibration mode of the winding machine in the finishing stage for winding processing. The winding machine executes the control signal in the final stage until the product integrity sensor determines that the product is completely wound. The integral value is calculated by the integrity calculation function. The function is I = ∫t0 to t1 f(x) dx, where I represents the integrity, f(x) is the function of the integrity with respect to time, and t0 and t1 represent the time of the winding stage. The start and end of the winding stage are set by the program, which includes a timer. If the winding result is incomplete, the control information is generated and transmitted to the motor execution unit through the transmitter, and the winding continues. After obtaining the complete winding product, the actual winding product final data is obtained in combination with the prediction model. The final winding process is statistically analyzed to obtain the actual result of the final data, and a data comparison chart is generated.

[0065] Assume that the prediction model outputs a stop advance of 5 meters in the final stage. First, obtain the stop advance data and fuse it with the preset path model information for the final stage. For example, the path model information indicates that the motor needs to decelerate according to a specific curve in the final stage. The fused control instruction combines the stop advance with the curve. For example, the initial speed of the winding motor is set to 1000 rpm, the target speed is 450 rpm, and the deceleration time is 5 seconds. The fused instruction is to smoothly reduce the speed to 450 rpm within 5 seconds when winding the last 5 meters.

[0066] In order to improve stability and accuracy, a smooth transition is adopted, such as using an S curve. The deceleration model can be calculated using the S curve in control theory. The formula can be simplified as V(t) = V0 + (Vt - V0) * (1 + exp(-k(t - t0)))^(-1), where V(t) is the speed at time t, V0 is the initial speed, which is 1000 rpm here, Vt is the target speed, which is 450 rpm here, and k is a constant that can be adjusted according to the speed of deceleration. Here, k=1, t0=5 seconds, which is used to control the shape and transition period of the S curve. V(t) = 1000 + (450-1000) * (1 + exp(-1*(t-5)))^(-1) is an S curve that describes the change of speed over time. The winding machine driver obtains this fused motor control instruction and drives it in combination with the current winding motor speed data. For example, by obtaining the speed through the encoder installed on the motor shaft, real-time speed feedback data can be obtained, which can be compared with the control curve. Set the preset timing parameters, such as collecting data every 10 milliseconds and setting the control cycle to 10 milliseconds. The optical fiber status sensor group and the optical fiber tension sensor group monitor the optical fiber diameter and tension in real time.

[0067] When the sensor detects that the change in fiber diameter at a certain moment exceeds the set threshold of 0.2 mm, or the fiber tension is greater than the set threshold of 5 Newtons, the fiber status monitoring module determines it as abnormal and shuts down. At this time, the fiber abnormal status data marking action is triggered, for example, the current data is marked as "abnormal diameter change is too large" or "abnormal tension is too large". These abnormal marks are identified by a pre-trained neural network model, which can be built based on a convolutional neural network or a recurrent neural network and trained with a large amount of historical data. For example, a neural network with a recognition accuracy of 98% has been trained. The training data set of the neural network is 10,000 samples. The neural network includes 1 input layer, 3 convolutional layers, 1 fully connected layer, and 1 output layer. Then further determine the specific abnormal mark, such as "abnormal diameter sudden increase", "abnormal diameter sudden decrease", "abnormal tension spike", etc. According to the specific abnormality determined, modify the mark in the record in the winding database. If the number of abnormal marks is greater than 0, for example, there are 3 "abnormal tension spikes", in order to compensate for the extra length requirement caused by the winding quality, the shutdown advance may be appropriately lengthened. Re-plan the operating parameters of the finishing control link according to the abnormal mark, for example, adjust the shutdown advance to 8 meters, adjust the parameters in the S curve formula, for example, adjust the formula V(t) = 1000 + (450-1000) * (1 + exp(-5*(t-5)))^(-1). A convolutional neural network based on 2 convolutional layers, 2 pooling layers, and 1 fully connected layer has been used to train a usable fiber finishing process parameter relationship model through training of historical production data, and the prediction accuracy has reached 99%. The convolutional neural network determines whether there is a fiber abnormal state data marking result and whether there is a product defect. If the judgment is "yes", secondary winding is performed, and the winding machine is controlled to proceed to the vicinity of the previously marked abnormal data for repair, such as winding more tightly or with a smaller wire diameter, so as to make up for the defect so that the final winding product meets the quality standards, or if the defect cannot be repaired, the product will be marked as defective. Otherwise, the secondary winding operation is not performed, and a prompt message is generated to the main control program. The control signal is transmitted to the motor drive unit of the winding machine using an industrial Ethernet protocol, such as PROFINET or EtherCAT. After receiving the signal, the drive unit controls the winding machine to perform the finishing stage according to the "winding mode" and "low vibration mode". If the winding mode is normal and it is a low vibration mode. The winding machine executes the control signal until the product integrity sensor determines that the product is completely wound, and calculates the integral value of the integrity calculation function. For example, when the integral value I reaches the set value of 100, it is considered that the product is completely wound, and the winding time is from 1 second to 10 seconds.The function of completeness with respect to time can be simplified as f(x)=10x, and the integral value is calculated, that is, I=∫1 to 10 f(x)dx, and I=495 is obtained, which meets the judgment condition of product winding completeness.

[0068] When the product is incomplete, continue winding. After obtaining the complete winding product, combine the prediction model and the actual finishing data, and generate a data comparison chart by comparing the differences between the two in terms of stop advance, motor speed, fiber diameter, tension, etc. The data statistics of the finishing winding process are generated and displayed on the display screen.

[0069] The above embodiments are only preferred technical solutions of the present invention and should not be regarded as limiting the present invention. The protection scope of the present invention shall be the technical solutions recorded in the claims, including equivalent replacement solutions of the technical features in the technical solutions recorded in the claims. That is, equivalent replacement improvements within this scope are also within the protection scope of the present invention.

Claims

1. A method for controlling the tension of optical fiber winding based on PLC, characterized in that The following steps are involved: S1: According to the preset total length of the optical fiber and the real-time winding length information fed back by the current encoder, the remaining length of the optical fiber is calculated by the length reduction model; S2: If the remaining length of the optical fiber is less than or equal to the preset winding trigger length threshold, the winding control unit switches from the normal winding state to the winding preparation state, and obtains a winding control unit signal that the winding control unit is about to perform the winding operation; S3: triggering the motor deceleration control module according to the finishing preparation state signal, using a preset deceleration curve control algorithm to smoothly reduce the motor speed, and determining that the motor running speed has reached the finishing winding speed value; S4: When the motor runs at a low speed and the length of the wound optical fiber reaches the value obtained by subtracting a safety margin from the final trigger length, the winding control unit obtains the current position information of the winding mechanism through a photoelectric sensor installed at a preset reference position, and determines the angle of the optical fiber relative to the preset reference position in the final winding stage; S5: According to the difference between the current angle of the real-time winding mechanism and the target angle, the winding control unit calculates the winding motor speed correction coefficient through the speed correction algorithm based on the position loop, obtains the control instruction for the fiber tail winding, and performs precise position control on the tail winding process; S6: If the output of the position control calculation model shows that the current angle of the winding mechanism has reached or exceeded the target angle, the winding control unit starts the preset neural network prediction model, and the neural network prediction model analyzes and learns from the historical operation data to obtain the advance amount information of the terminal parking position, and determines the amount of stop signal that needs to be issued in advance in the terminal control link; S7: Based on the stop advance of the finishing control link given by the prediction model and the preset finishing path model, the PLC outputs a stop signal to the driver, causing the winding motor to stop in a low-vibration deceleration mode to avoid fiber breakage and obtain a complete winding product.

2. The method for controlling the tension of optical fiber winding based on PLC according to claim 1, characterized in that: The sub-steps of S1 are: S2.1: If the optical fiber surplus length data value is less than or equal to the preset tail trigger threshold length value; S2.2: The winding control unit state data is updated to 1; S2.3: According to the winding control unit state data, when the winding control unit state data is 1, the winding control unit switches to the finishing preparation state; S2.4: The winding control unit obtains the finishing preparation state data; S2.5: The winding control unit enters the finishing preparation state; S2.6: The winding control unit generates the pre-trigger signal data for the closing operation; S2.7: The winding control unit pre-trigger signal data according to the closing operation; S2.8: The winding control unit runs a self-test on the equipment self-test module to obtain normal signal data of the equipment self-test; S2.9: Use the normal signal data of the equipment self-test as the precondition for the final closing operation; S2.10: The tailing signal is triggered and output, and the winding equipment performs the optical fiber tailing operation.

3. The method for controlling the optical fiber winding tension based on PLC according to claim 1, characterized in that: The sub-steps of S2 are: S2.1: If the optical fiber surplus length data value is less than or equal to the preset tail trigger threshold length value; S2.2: The winding control unit state data is updated to 1; S2.3: According to the winding control unit state data, when the winding control unit state data is 1, the winding control unit switches to the finishing preparation state; S2.4: The winding control unit obtains the finishing preparation status data.

4. The method for controlling the tension of optical fiber winding based on PLC according to claim 1, characterized in that: The sub-steps of S3 are: S3.1: The trigger generates a high level output according to the signal of the end preparation state. After the motor controller receives the high level, it obtains the information and starts the control; S3.2: The starting motor deceleration control module processes the end preparation state signal, transmits it to the programmable logic controller through the bus, and obtains the instruction data set; S3.3: Obtain the function relationship expression y=f(x) of the preset deceleration curve by table lookup method, where the independent variable x represents time and the dependent variable y represents motor speed, and obtain the expected speed set of the motor; S3.4: Obtain the real-time motor speed information measured by the encoder and transmit it to the data processor through the serial port to obtain the actual running speed value; S3.5: The Kalman filter algorithm is used to reduce the noise of the actual running speed value of the motor, and the actual running speed value after noise reduction is connected to the PID controller as a feedback input quantity, and the PID parameters are adjusted iteratively to obtain the optimal PID control instruction; S3.6: The PID controller combines the motor speed value obtained by the preset deceleration curve with the actual running speed value after noise reduction. If the difference between the two is greater than the set value, the alarm module is triggered to issue an early warning; S3.7: A prediction model is established using the BP neural network algorithm. The prediction model is trained using the actual operating speed set obtained through PID control. It is judged that the predicted speed value meets the final winding speed value and a stop signal is output.

5. The method for controlling the tension of optical fiber winding based on PLC according to claim 1, characterized in that: The sub-steps of S4 are: S4.1: The speed is recorded periodically by the encoder installed on the motor and a data stream is generated in combination with the time parameter; S4.2: This data flows through the data buffer pool for temporary storage; S4.3: The neural network prediction model adjusts its own parameters in real time through a self-learning process based on historical operating rate data and establishes an operating rate determination benchmark; S4.4: The model compares the current data with the judgment benchmark obtained through learning, determines the operation rate value and defines different working stages for different values. The operation rate less than a certain value is defined as the low-speed operation stage; S4.5: The running status value enters the judgment reference data stream as a status label for subsequent analysis and processing; S4.6: Calculate the number of winding turns according to the preset parameters of the optical fiber, set the initial turn value, combine the information collected by the photoelectric sensor to obtain the final turn value and the circumference of the winding reel, multiply them to obtain the total length at the current moment, and then subtract them to obtain the actual remaining length; S4.7: Subtract the tail trigger length setting value from the safety margin setting value to calculate the comparison length; S4.8: Using the preset rules to determine the actual remaining length and the comparison length value, to obtain a determination result of whether the remaining length is within the end trigger range; S4.9: If the length falls within the range, a tail trigger instruction is generated; S4.10: The ending trigger instruction is obtained according to the preset rules. The control logic of the ending stage parses the data packet in the instruction to obtain the control parameters and performs data classification processing, and distributes the classified control instructions to the designated device; S4.11: When the photoelectric sensor detects and obtains trigger information, the trigger information is stored in the buffer area; S4.12: A photoelectric sensor installed at a preset reference position of the winding device detects a periodic rotation signal of the winding disk and collects a rotation data sequence; S4.13: The photoelectric sensor converts the position data into an analog electrical signal of a period of one rotation of the winding device; S4.14: Perform frequency domain analysis on the acquired analog electrical signal using Fourier transform to obtain harmonic components and main frequency; S4.15: Calculate the total length of the optical fiber and the number of turns of the device, calculate the two to obtain the corresponding length information of each turn, and divide the remaining length of the optical fiber by the length of each turn to obtain the quotient and remainder; S4.16: The remainder value is multiplied by the duration corresponding to each circumference and the phase corresponding to each circumference to obtain the converted angle value; S4.17: Obtain the total phase angle and convert the remainder into radians based on the specified reference point; S4.18: A numerical calculation model is pre-set, and the numerical calculation model collects the numerical value after Fourier transformation and clock data in real time as input data, and uses differential calculation to obtain the numerical change over a period of time; S4.19: Time series acquisition All data obtained from S4.1-S4.18 above are fused; S4.20: Compare the photoelectric sensor information with the reference time series through the time window to obtain the time correlation, and then combine the mechanical structure motion amplitude information corresponding to the reference time series to obtain the phase offset value, so as to obtain the time parameter; S4.21: Calculate the product of the time value and the angular velocity value to obtain the total angle; S4.22: According to the time parameter, the instantaneous angular velocity of the current motor and the corresponding optical fiber length value are determined; S4.23: The system updates the trigger length threshold of the closing stage in real time according to the fiber length, and uses the fiber type data in the established database as the judgment object; S4.24: Obtain the final parameter value of each material, and send the final parameter value to the motion controller through the data interface; S4.25: Conduct a comprehensive comparative analysis based on the mechanical structure characteristic values ​​of optical fiber winding equipment; S4.26: Determine the actual position deviation value through the output value of the photoelectric sensor and the internal setting control parameters; S4.27: The recurrent neural network predicts the winding position information for the next period and obtains multiple future time prediction points; S4.28: Subtract the predicted point from the current value to obtain the position change trend and change speed, and use the trend data as the basis for feedback control data in real time; S4.29: The motor operating parameters are adjusted by the fuzzy control algorithm according to the deviation value, and the photoelectric sensor monitors a preset time value to synchronize the time and value data and store them in the historical database; S4.30: The historical database classifies and stores data of different storage types, thereby achieving speed adjustment.

6. The method for controlling the tension of optical fiber winding based on PLC according to claim 1, characterized in that: The sub-steps of S5 are: S5.1: Obtain current angle information of the winding mechanism and set target angle information according to real-time detection; S5.2: Calculate the angle difference according to the current angle information and the target angle information. When the angle difference exceeds the preset range, enter the early warning database; S5.3: Read the angle difference from the warning database. If the angle difference is lower than the preset lower limit, skip the speed correction step. S5.4: If the angle difference is higher than the preset upper limit range, an alarm message corresponding to the angle difference is generated and the display module is notified to display the message, and the control instruction generation is paused. At the same time, the current angle and the target angle are read from the warning database; S5.5: Calculate the angle difference by a control algorithm based on a position loop to obtain a speed correction coefficient; S5.6: Correct the current output torque of the winding motor according to the speed correction coefficient to obtain corrected real-time torque control information; S5.7: adaptively compensate the real-time torque control information according to the preset target length of the optical fiber winding to obtain a final control instruction; S5.8: Establish an environment perception database for the real-time winding environment, and send the final control instruction to the winding control actuator. The actuator receives the final control instruction and controls the winding motor to perform the optical fiber finishing winding work; S5.9: If the final state parameter in the final control command shows stop, the actuator is suspended from winding, and the control command update is suspended, then the working data update information in the environmental perception database is obtained, the new winding target length information is obtained, and the corresponding control command information is read from S5.5, and the command is resent; S5.10: If the final state parameter in the final control instruction does not indicate stop, continue with this step; S5.11: Perform online detection according to the real-time data information library of the winding mechanism execution status, determine the movement status of the final winding work and save it.

7. The method for controlling the tension of optical fiber winding based on PLC according to claim 1, characterized in that: The sub-steps of S6 are: S6.1: Obtain current angle information of the winding mechanism from the winding mechanism, and determine the difference between the current angle information and the target angle information; S6.2: If the difference between the current angle information and the target angle information is greater than the set value, the winding control unit outputs a signal to the neural network prediction model; S6.3: The neural network prediction model obtains the signal of the winding control unit, and retrieves the historical operation data from the database for analysis; S6.4: According to the analysis result, the advance amount information of the end parking position is obtained; S6.5: Determine the amount of advance of the stop signal issued by the terminal control according to the advance amount information of the terminal parking position; S6.6: The winding control unit receives the stop signal and controls the winding mechanism to perform corresponding operations; S6.7: The database receives and stores the angle information, signal quantity and winding mechanism information.

8. The method for controlling the tension of optical fiber winding based on PLC according to claim 1, characterized in that: The sub-steps of S7 are: S7.1: Obtain the stop advance data of the final stage output by the prediction model, and obtain the motor control command by fusing the stop advance data and the preset final stage path model information; S7.2: The winding machine driver obtains the fused motor control command. According to the current rotational speed data of the winding motor, the driver controls the winding motor to run to the target speed, and the target speed is lower than half of the initial winding speed; S7.3: Through a preset timing sequence, the timing parameters are from the database. The optical fiber status sensor group obtains the optical fiber diameter data during operation, and at the same time, the optical fiber tension sensor group obtains the real-time tension data. The operation state is estimated using historical moment data. When the change amplitude of the optical fiber diameter is greater than the set threshold within at least one moment, or when the optical fiber tension is greater than the threshold within at least one moment, the optical fiber status monitoring module makes a shutdown judgment, triggers the action of marking the optical fiber abnormal state data, and the specific abnormal mark is determined after being recognized by the neural network, and then the result of marking the optical fiber abnormal state data is recorded; S7.4: In the final stage, collect the result of marking the optical fiber abnormal state data obtained through step S7.3, and analyze whether the number of all abnormal marks is 0; S7.5: In the final stage, use a convolutional neural network modeled according to the relationship of optical fiber finishing process parameters to judge whether secondary winding is required; If there is a result of marking the optical fiber abnormal state data and there are product defects, enter the secondary winding control link. Through the defect type and position data, and combined with the defect elimination strategy, plan a new set of winding control instructions, and calculate and generate the actual motor operation control instructions according to the control instruction set and the product defect position coordinate values; Otherwise, no secondary winding operation is performed; S7.6: The winding machine driver obtains the control signal after secondary winding; S7.7: Through the signal transmission device, it is transmitted to the winding machine motor drive unit using industrial Ethernet; S7.8: The drive unit determines the winding mode and low-vibration mode of the winding machine in the final stage for winding processing; S7.9: The winding machine executes this control signal in the final stage until the product integrity sensor determines that the product winding is complete, and the integral value is calculated through the integrity calculation function. The function is I = ∫ from t0 to t1 f(x)dx, where I represents the integrity, f(x) is the function of integrity with respect to time, and t0 and t1 represent the time of the winding stage; If the winding result is incomplete, generate control information to reach the motor execution unit through the transmitter, and the winding continues; S7.10: After obtaining the completely wound product, combine the prediction model to obtain the actual winding product finishing data; S7.11: Conduct data statistics on the process of the finishing winding to obtain the actual result of the finishing data, and generate a data comparison chart; S7.12: Data storage and analysis: Store the data in the final stage and use it for subsequent analysis and improvement to ensure that each winding process can learn from historical data, continuously optimize the control algorithm, and improve the product quality and production efficiency.

9. The method for controlling the tension of optical fiber winding based on PLC according to claim 1, characterized in that: The sub-steps of step S7.4 are: S7.4.1: If the number of abnormal marks is greater than 0, then based on the abnormal marks, re-plan the parameters of the finishing control link; S7.4.2: Generate a control signal based on the stop advance amount of the tail control link obtained in S7.4.1 and send it to the motor control link, and the motor control link adjusts the control curve; S7.4.3: If the number of abnormal marks is 0, the control link parameter regeneration mechanism will not be triggered, the control link parameters will remain unchanged as in S7.4.1, and subsequent operations will continue.

10. A PLC-based optical fiber winding tension control system, characterized in that: A PLC-based optical fiber winding tension control method according to any one of claims 1 to 9 is adopted.

Images