Search

CN-122009915-A - Intelligent self-adaptive yarn tension control system and method

CN122009915ACN 122009915 ACN122009915 ACN 122009915ACN-122009915-A

Abstract

The application relates to the field of tension control, in particular to an intelligent self-adaptive yarn tension control system and method, which comprises a data acquisition module, an adjusting module, a prediction module and a control module, wherein the data acquisition module is used for acquiring yarn state data, the adjusting module is used for adjusting yarn tension, the prediction module predicts yarn breakage risk based on time sequence data, the control module receives data of the data acquisition module and the prediction module to control the operation of the adjusting module, the prediction module uses a time sequence mode map of yarn breakage history to establish breakage risk so as to output a breakage probability average value and risk position information in real time, so that tension adjustment control is performed on a yarn position with high possibility of yarn breakage in advance and accurately, and the sudden yarn breakage possibility is reduced.

Inventors

  • DAI YAQIN
  • CHE WEIMIN
  • LI JUTAO
  • DONG HAIQIANG

Assignees

  • 嘉兴驰逸智能科技有限公司

Dates

Publication Date
20260512
Application Date
20260317

Claims (10)

  1. 1. An intelligent adaptive yarn tension control system, characterized by comprising: the data acquisition module is used for acquiring yarn state data; The adjusting module is used for adjusting the tension of the yarns; a prediction module that predicts a yarn breakage risk based on the time-series data; the control module receives the data of the data acquisition module and the prediction module to control the operation of the adjustment module; the prediction module uses a history record of yarn breakage to establish a time sequence mode mapping of breakage risk so as to output a breakage probability average value and risk position information in real time.
  2. 2. An intelligent self-adaptive yarn tension control method based on the operation of the intelligent self-adaptive yarn tension control system as claimed in claim 1 is characterized by comprising the following steps: s1, acquiring multichannel original time sequence data in real time, adding a hardware time stamp to each frame of original time sequence data, marking a data quality state, and generating an environment mutation trigger signal when the change rate of environment parameters exceeds a preset threshold value; s2, carrying out online drift detection and dynamic calibration on original time sequence data, identifying zero drift of a tension sensor and a visual lens fogging event, carrying out real-time compensation calculation on the drift data, carrying out virtual feature filling on disabling visual data, and constructing an enhanced time sequence tensor containing an environment transient feature vector; S3, inputting the enhanced time sequence tensor into a dual-path LSTM framework, extracting a basic tension fluctuation mode by a main path, generating an environment disturbance compensation vector by an anti-drift adapter path according to a drift detection state, and outputting fracture probability mean value, prediction variance and risk position information; s4, dynamically adjusting a fracture alarm threshold based on a prediction variance, lifting the threshold to inhibit false alarm when the confidence level of the model is low, reducing the threshold to prevent false alarm when a drift mark and tension peak composite condition is detected, and generating a differential control instruction with a labeling source according to a root cause decision matrix; s5, issuing and executing a differential control instruction, collecting feedback data after execution to verify the intervention effect, counting false alarm/missing alarm events and triggering a model self-updating signal.
  3. 3. The method for intelligently and adaptively controlling the yarn tension according to claim 2, wherein the multichannel original time sequence data in S1 comprises tension, visual images and temperature and humidity; Enabling redundant acquisition of the main and standby double-channel strain gauges for tension data, and marking the data of the corresponding period as an uncertain state when the double-channel difference exceeds 3cN and lasts for 200 milliseconds; When the absolute value of the humidity change rate is more than 10% RH/second or the absolute value of the temperature change rate is more than 2 ℃/second, generating an environment abrupt change trigger signal and packaging the environment abrupt change trigger signal into the head part of the original time sequence data frame.
  4. 4. The method for intelligent self-adaptive yarn tension control according to claim 2, wherein a window drift detection algorithm is adopted in S2 to maintain a statistical test window, and when an environmental mutation trigger signal is activated and the average difference of the tension of adjacent windows exceeds 5cN, the environmental induced zero drift is determined.
  5. 5. The method of claim 2, wherein S2 calculates a Laplace variance sharpness index for the visual data stream, and when sharpness is less than a fogging threshold and a humidity change rate is greater than 5% RH/sec, the method determines a lens fogging event and marks the visual data as invalid.
  6. 6. The method for intelligent self-adaptive yarn tension control according to claim 2, wherein S2 calls an offline calibrated temperature-humidity-drift compensation table to perform online zero regression compensation on the drift tension value, and virtual visual features are generated by combining last valid visual frames before fogging with an autoregressive model extrapolation, and low confidence weight is given.
  7. 7. The method of claim 2, wherein S3 fuses the main path output and the compensation vector by a dynamic weight, the dynamic weight is 0.5 when the drift detection mark is activated and 0.1 when the drift detection mark is not activated, and the conditional activation of the model parameters is realized.
  8. 8. The method for intelligent self-adaptive yarn tension control according to claim 2, wherein S4 sets a basic alarm threshold value to be 0.7, automatically raises the threshold value to be 0.85 when the confidence level of the model is low to enter a conservative mode, and automatically lowers the threshold value to be 0.55 when a composite condition that a drift mark and a tension peak value exceed 90cN is detected to enter an early warning mode.
  9. 9. The method of claim 8, wherein the model confidence level is classified according to a high confidence level when the prediction variance is less than 0.01, a medium confidence level between 0.01 and 0.05, and a low confidence level when the prediction variance is greater than or equal to 0.05, and the priority arbitration is implemented through a priority value field in the instruction frame, wherein the value range is 1 to 10.
  10. 10. The intelligent self-adaptive yarn tension control method according to claim 2, wherein S5 verifies whether the drift flag is cleared within 5 seconds for the calibration mode instruction, and if not, the calibration is judged to be failed and the alarm is updated; Recalculating the average value of the fracture probability of the data 3 seconds after the acquisition and execution of the intervention mode instruction, and judging that the intervention is invalid and triggering a secondary response if the risk reduction is less than 0.2; And maintaining a false alarm counter and a false alarm counter, increasing the counterloss weight of the model field on line when the false alarm rate exceeds 5%, and generating a model architecture upgrading request to trigger off-line retraining when the anti-interference gain is continuously lower.

Description

Intelligent self-adaptive yarn tension control system and method Technical Field The application relates to the field of tension control, in particular to an intelligent self-adaptive yarn tension control system and method. Background The yarn tension needs to be precisely controlled to reduce the possibility of yarn breakage, to reduce the stop frequency as much as possible, and to reduce the waste of yarn and grey cloth, especially in the weaving process of high-density fabrics, for which the yarn tension needs to be precisely controlled. Some existing tension control technologies mainly depend on fixed parameter models, such as preset PID gain and a Kalman filter Q/R matrix, can refer to an intelligent yarn low tension control system with the publication number of CN120905903A approximately, realize low tension stable control through preset PID gain and filtering parameters, have certain applicability in middle-low speed, middle-low density yarn production, but still have obvious technical short plates in complex production scenes. However, during high density yarn production, yarn breaks often result from accumulated fatigue or implicit defects, such as tension fluctuations several times exceeding a threshold value, internal microcracks within the yarn, and the like. The specific reasons are from analysis of yarn mechanical properties, and broken yarns are the results of local damage of macromolecules in the fiber, stress concentration at structural defects and plastic deformation accumulation. During repeated stretching and bending cycles, the yarn gradually has loose structure and reduced elastic recovery rate, and finally is broken due to fatigue. And the hidden defects such as microcracks, fiber entanglement, impurity inclusion and the like in the yarn are key causes of stress concentration. The real-time tension value is collected by the tension sensor, the defects on the surface of the yarn are detected by the industrial camera, but the existing tension sensor can only capture the instantaneous tension peak value and cannot identify the time sequence accumulation characteristics of tension fluctuation, such as repeated small-amplitude super-threshold fluctuation in a short time, and the industrial camera can only detect dominant defects such as surface hairiness, uneven thickness and the like and cannot detect hidden defects such as internal microcracks. Even if perspective detection technologies such as X-rays and ultrasonic waves are adopted, the problems of insufficient resolution, fuzzy imaging under high-speed movement, high equipment cost, strict radiation protection requirements and the like exist, the method is difficult to apply on a large scale on a production line, the monitoring dimension is limited to an instant state, and the sudden yarn breakage probability is still higher. Disclosure of Invention In order to reduce the sudden yarn breakage probability, the application provides an intelligent self-adaptive yarn tension control system and method. In a first aspect, the present application provides an intelligent adaptive yarn tension control system according to the following technical solution. An intelligent adaptive yarn tension control system, comprising: the data acquisition module is used for acquiring yarn state data; The adjusting module is used for adjusting the tension of the yarns; a prediction module that predicts a yarn breakage risk based on the time-series data; the control module receives the data of the data acquisition module and the prediction module to control the operation of the adjustment module; the prediction module uses a history record of yarn breakage to establish a time sequence mode mapping of breakage risk so as to output a breakage probability average value and risk position information in real time. In a second aspect, the present application provides an intelligent adaptive yarn tension control method, which adopts the following technical scheme. An intelligent self-adaptive yarn tension control method, which uses the intelligent self-adaptive yarn tension control system, specifically comprises the following steps: s1, acquiring multichannel original time sequence data in real time, adding a hardware time stamp to each frame of original time sequence data, marking a data quality state, and generating an environment mutation trigger signal when the change rate of environment parameters exceeds a preset threshold value; s2, carrying out online drift detection and dynamic calibration on original time sequence data, identifying zero drift of a tension sensor and a visual lens fogging event, carrying out real-time compensation calculation on the drift data, carrying out virtual feature filling on disabling visual data, and constructing an enhanced time sequence tensor containing an environment transient feature vector; S3, inputting the enhanced time sequence tensor into a dual-path LSTM framework, extracting a basic tension fluctuation mode by a main path