Disclosed are a tension self-adaptive control method and system for a disc root braiding machine.

Abstract

The present application belongs to the field of automatic control of sealing material production equipment, and particularly relates to a tension self-adaptive control method and system for a packing braider, which comprises the following steps: acquiring a real-time rotating speed of a main shaft and a continuous tension signal; constructing an adaptive band-pass filter according to the real-time rotating speed of the main shaft; calculating a local information entropy deviation and an envelope mean value of a high-frequency vibration component in a sliding window; extracting a current calculation period representative value of a low-frequency drift component and calculating a drift gradient, accumulating a drift amount and judging a direction continuity; superimposing a feedforward compensation instruction and a feedback control instruction generated based on a feedback proportional gain and a normalized tension error, and performing a nonlinear smoothing processing to obtain a final tension control instruction. Through the cooperation of dynamic decoupling, credible feedforward and robust feedback, the present application effectively improves the stability and anti-disturbance ability of the tension control of the packing braider.

Description

Technical Field

[0001] This invention relates to the field of automatic control of sealing material production equipment. In particular, it relates to a tension adaptive control method and system for packing braiding machines. Background Technology

[0002] Packing (also known as sealing filler) is a high-performance sealing material made from industrial specialty fibers such as aramid, polytetrafluoroethylene (PTFE), and graphite through a weaving process. It is widely used in dynamic and static sealing applications such as pumps and valves in various industrial equipment. Packing weaving machines are specialized industrial weaving equipment used to weave these industrial specialty fiber raw materials into finished packing products. They fall under the category of sealing material production equipment. Their processing targets are industrial specialty fibers such as aramid, PTFE, and graphite, rather than ordinary cotton, linen, silk, or wool fibers, and are fundamentally different from general textile machinery used for civilian fabric weaving.

[0003] During the packing braiding process, the tension of the yarn (i.e., the fiber strands used to braid the packing, such as aramid yarn, polytetrafluoroethylene yarn, graphite yarn, etc., hereinafter the same) is affected by multiple factors such as the spindle mechanical cycle, roll diameter consumption, and random disturbances, exhibiting a complex fluctuation characteristic of high-frequency vibration and low-frequency drift. To achieve stable control, existing methods typically decompose the continuous tension signal into high-frequency vibration components and low-frequency drift components, generating feedforward compensation commands for the high-frequency components and applying feedback regulation to the low-frequency components. Common decomposition methods rely on filters with a fixed center frequency, extracting high-frequency components through bandpass filtering, and then obtaining low-frequency signals through differential or low-pass filtering, thereby generating corresponding control actions.

[0004] However, in actual operation, the spindle speed is not constant due to power grid fluctuations, load changes, and other factors. A filter with a fixed center frequency cannot synchronously track the resulting weaving fundamental frequency offset, leading to passband mismatch, spectral leakage, and phase distortion. This causes high-frequency and low-frequency components to overlap, reducing decoupling accuracy. Furthermore, conventional solutions often directly calculate feedforward compensation based on the instantaneous envelope or local energy of the high-frequency vibration components, lacking judgment on whether the vibration pattern originates from normal periodic weaving. Once non-periodic disturbances such as yarn knots, sensor spikes, or mechanical shocks occur, these disturbances are also included in the feedforward path and may be amplified into actual tension fluctuations. In the feedback channel, the adjustment of the proportional gain often relies solely on the instantaneous difference of low-frequency components between adjacent control cycles, making it difficult to distinguish between the continuous gradual changes caused by yarn diameter loss and the occasional fluctuations caused by short-term friction or random jitter. This results in frequent jumps in the gain parameter under disturbances, weakening the stability margin of the closed-loop system. Furthermore, when the superimposed control commands approach or exceed the physical saturation boundary of the actuator, the direct truncation limiting method is prone to generating command step, which impacts the actuator and causes sudden changes in yarn tension. Summary of the Invention

[0005] To address the problems of inaccurate dynamic decoupling, susceptibility of feedforward to aperiodic disturbances, poor stability of feedback parameters, and discontinuous boundary output in the tension control of packing braiding machines under multi-source time-varying interference, this invention provides solutions in the following aspects.

[0006] In a first aspect, this invention discloses a tension adaptive control method for a packing braiding machine, comprising: acquiring the real-time spindle speed and continuous tension signal, and acquiring the upper limit and lower limit of the saturated output of the actuator; determining the interlacing dynamic fundamental frequency based on the real-time spindle speed, constructing an adaptive bandpass filter with the interlacing dynamic fundamental frequency as the center frequency, decomposing the continuous tension signal to obtain high-frequency vibration components and low-frequency drift components; calculating the local information entropy deviation and envelope mean of the high-frequency vibration components within a sliding window; when the local information entropy deviation is within a preset allowable range of information entropy deviation, generating a feedforward compensation command based on the envelope mean and the real-time spindle speed; otherwise, limiting the increment of the feedforward compensation command; wherein, the preset allowable range of information entropy deviation is that the local information entropy deviation does not exceed a preset upper limit of information entropy deviation; extracting low-frequency components in units of a preset calculation period. The drift component is represented by the current calculation cycle value. The drift gradient is calculated based on the representative values ​​of the current and previous calculation cycles. The drift gradient is accumulated to obtain the drift accumulation amount. The continuity of the drift direction is judged. When the drift accumulation amount exceeds the preset drift threshold and the drift direction is continuous, the preset feedback proportional gain is corrected. Otherwise, the feedback proportional gain before correction is maintained. The feedforward compensation command is superimposed with the feedback control command generated based on the feedback proportional gain and normalized tension error. Nonlinear smoothing is performed according to the upper and lower limits of the actuator's saturated output to obtain the final tension control command. The nonlinear smoothing is as follows: when the superimposed command value is between the upper and lower limits of the saturated output, it remains unchanged. When it exceeds the limits, a continuously differentiable saturation function is used to compress the command value to between the upper and lower limits of the saturated output.

[0007] Preferably, determining the interlacing dynamic fundamental frequency includes: calculating the interlacing dynamic fundamental frequency based on a preset transmission ratio coefficient and the real-time spindle speed; the transmission ratio coefficient is the mechanical conversion ratio from the spindle speed to the interlacing frequency.

[0008] Preferably, the step of constructing an adaptive bandpass filter with the interleaved dynamic fundamental frequency as the center frequency includes: presetting the bandwidth isolation tolerance width; constructing an adaptive bandpass filter with the difference between the interleaved dynamic fundamental frequency and the bandwidth isolation tolerance width as the lower cutoff frequency and the sum of the interleaved dynamic fundamental frequency and the bandwidth isolation tolerance width as the upper cutoff frequency; obtaining the group delay of the adaptive bandpass filter; and performing phase compensation on the high-frequency vibration components based on the group delay before generating the feedforward compensation command.

[0009] Preferably, the calculation of the local information entropy deviation and envelope mean of the high-frequency vibration component includes: calculating the mean of the absolute amplitude within a sliding window of the high-frequency vibration component as the envelope mean; within the sliding window, dividing the amplitude of the high-frequency vibration component into equal-width bins according to a preset number of bins, statistically analyzing the sample proportion within each bin as the amplitude distribution, calculating the local information entropy according to the Shannon information entropy formula based on the amplitude distribution; obtaining the reference information entropy; and using the absolute value of the difference between the local information entropy and the reference information entropy as the local information entropy deviation.

[0010] Preferably, the acquisition of reference information entropy includes: acquiring the existing reference information entropy for the current period; calculating the local information entropy deviation for the current period based on the local information entropy for the current period and the existing reference information entropy for the current period; updating the local information entropy for the current period to a new reference information entropy when the sliding window meets the preset stability conditions; otherwise, keeping the existing reference information entropy for the current period unchanged; the preset stability conditions include: the real-time rotational speed of the spindle within the sliding window is effective, the continuous tension signal does not have missing data or exceed the preset effective signal range, the local information entropy deviation for the current period is within the preset allowable range of information entropy deviation, and the envelope change of the high-frequency vibration component does not manifest as an isolated impact; wherein, an isolated impact is defined as: the change in the mean envelope value within the sliding window relative to the mean envelope value of the previous sliding window exceeds the product of the preset impact judgment multiple and the mean envelope value of the previous sliding window, and the change falls back to within the preset fallback ratio of the mean envelope value of the previous sliding window in the next sliding window.

[0011] Preferably, the generation of the feedforward compensation command includes: generating a feedforward damping compensation factor based on the envelope mean, the real-time spindle speed, a preset damping mapping gain, and a speed sensitivity index, by multiplying the damping mapping gain by the product of the envelope mean and the speed sensitivity index of the real-time spindle speed raised to the power of the speed sensitivity index; the speed sensitivity index is used to characterize the nonlinear influence of spindle speed changes on the feedforward damping compensation strength; and linearly mapping the feedforward damping compensation factor using a preset feedforward mapping gain and a preset bias to obtain the feedforward compensation command; wherein, the linear mapping is the product of the feedforward mapping gain and the feedforward damping compensation factor plus the preset bias.

[0012] Preferably, the limitation of the increment of the feedforward compensation instruction includes: when the local information entropy deviation exceeds the preset allowable range of information entropy deviation, using the increment of the feedforward compensation instruction of the previous calculation cycle as the increment of the feedforward compensation instruction of the current calculation cycle, or reducing the increment of the feedforward compensation instruction of the current calculation cycle according to a preset attenuation ratio.

[0013] Preferably, the determination of drift direction continuity includes: counting the symbols of the drift gradient within a preset number of consecutive calculation cycles; when the proportion of the number of symbols in the main direction to the number of valid symbols reaches a preset direction proportion, and the number of symbol reversals does not exceed a preset number of reversals, the drift direction is determined to be continuous.

[0014] Preferably, the step of correcting the preset feedback proportional gain includes: calculating the updated feedback proportional gain based on the feedback proportional gain, the preset trend compensation convergence factor, and the drift gradient; the updated feedback proportional gain is the feedback proportional gain before correction plus the product of the trend compensation convergence factor and the drift gradient.

[0015] In a second aspect, the present invention discloses a tension adaptive control system for a packing braiding machine, comprising: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the tension adaptive control method for the packing braiding machine described in any one of the claims is implemented.

[0016] The beneficial effects of this invention are: 1. Calculate the interlaced dynamic fundamental frequency based on the real-time spindle speed and construct an adaptive bandpass filter centered on it. This synchronizes the frequency band division of high-frequency interlaced vibration and low-frequency roll diameter drift in the tension signal with the real-time mechanical vibration frequency, reducing spectral leakage and phase distortion caused by speed drift. Compensate for the filter group delay to align the phase of the extracted high-frequency components with the actuator compensation phase. Quality labeling of tension samples prevents outliers from being mixed into subsequent information entropy reference updates and low-frequency representative value calculations, thus providing a purer and more accurate component source for feedforward and feedback.

[0017] 2. For high-frequency components, calculate the local information entropy within a sliding window. Determine whether the vibration energy distribution conforms to the periodic interlacing characteristics by the deviation from the information entropy reference value of the stable window. When the information entropy deviation is within the allowable range, generate a feedforward damping compensation factor based on the envelope mean and map it to a feedforward command, so that the feedforward compensation only applies to identifiable stable interlacing vibrations. When the deviation exceeds the limit, limit the feedforward increment and pause the information entropy reference update to avoid non-periodic disturbances such as yarn knots and sensor spikes being mistakenly amplified as damping commands, thereby reducing additional tension fluctuations caused by miscompensation.

[0018] 3. Multi-cycle representative values ​​are extracted from low-frequency drift components, and the drift gradient is calculated. The continuity of the cumulative drift amount and gradient direction is used as a determinant; feedback proportional gain correction is triggered only when the cumulative amount exceeds a threshold and the direction remains consistent. This ensures that the gain adjustment corresponds to a continuous, gradual trend in parameters such as convolutional diameter consumption, preventing frequent parameter jumps and closed-loop oscillations caused by short-term single-cycle disturbances. The superimposed control command undergoes nonlinear smooth mapping based on the actuator saturation boundary, ensuring that the command gradually approaches the physical boundary rather than abruptly truncating it, reducing actuator impact and maintaining stable tension control output. Attached Figure Description

[0019] Figure 1 This is a flowchart of the tension adaptive control method for a packing braiding machine according to an embodiment of the present invention. Detailed Implementation

[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0021] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0022] Reference Figure 1 The tension adaptive control method for packing braiding machines includes steps S1-S5, as detailed below: S1: Obtain the real-time spindle speed and continuous tension signal, and obtain the upper limit and lower limit of the actuator's saturated output.

[0023] In one embodiment, during the operation of the packing braiding machine, the spindle rotation drives the braiding action, and the yarn tension changes dynamically with the braiding process.

[0024] The real-time spindle speed and continuous tension signal are two fundamental physical quantities that reflect the state of the weaving process.

[0025] In this step, the real-time spindle speed is read from the main control system of the packing braiding machine and recorded as follows: The unit is revolutions per minute; continuous tension signals are collected from tension sensors installed along the yarn path to form a continuous tension time series, denoted as... The unit is Newton.

[0026] Data acquisition is performed with a fixed control cycle, which is set according to the running speed of the knitting machine and the real-time control requirements, for example, 20ms.

[0027] The acquisition of rotational speed from the main control system and the collection of signals from the tension sensor are achieved through industrial fieldbus communication or analog input channels. This data acquisition and transmission technology is well-known and will not be described in detail here.

[0028] Through the above acquisition process, a pair of the latest data is obtained in each control cycle. and Samples are used to create real-time data records.

[0029] S2: Determine the interlacing dynamic fundamental frequency based on the real-time spindle speed, construct an adaptive bandpass filter with the interlacing dynamic fundamental frequency as the center frequency, decompose the continuous tension signal to obtain high-frequency vibration components and low-frequency drift components.

[0030] In one embodiment, during the actual operation of the packing braiding machine, the spindle speed is not absolutely constant due to factors such as power grid fluctuations and changes in mechanical load. When the filter with a fixed center frequency decomposes the continuous tension signal, spectral leakage and phase distortion will occur due to the offset of the actual interlacing fundamental frequency, resulting in the superposition of high-frequency interlacing vibration components and low-frequency roll diameter drift components.

[0031] Therefore, in this embodiment of the invention, the interlacing dynamic fundamental frequency at the current moment is calculated based on the real-time spindle speed to ensure that the center frequency of the frequency band division follows the actual weaving conditions.

[0032] The interlacing dynamic fundamental frequency is determined by the real-time spindle speed and the transmission ratio coefficient, and the calculation formula is as follows: In the formula, The interlacing dynamic fundamental frequency, measured in Hertz, characterizes the theoretical vibrational characteristic frequency generated by the yarn during mechanical interlacing. The real-time spindle speed, measured in revolutions per minute, is obtained from the main control system of the packing braiding machine. The transmission ratio coefficient is dimensionless and represents the mechanical conversion ratio from the spindle speed to the weaving frequency. It is provided by the gear ratio or mechanical structure parameters of the braiding machine's transmission chain, or it can be determined based on the equipment calibration data. In this embodiment, an empirical value of 0.02 is taken as the preset parameter, which can also be adjusted by the implementer according to the specific equipment transmission structure.

[0033] Since it is necessary to separate the high-frequency vibrations that are highly correlated with the mechanical interlacing period and the low-frequency drift caused by slowly varying factors such as roll diameter loss in the continuous tension signal, this step constructs an adaptive bandpass filter centered on the interlacing dynamic fundamental frequency.

[0034] Lower cutoff frequency of bandpass filter and upper cutoff frequency Set them to: , In the formula, It represents the bandwidth of the frequency isolation tolerance, measured in Hertz, and is used to limit the frequency range centered on the dynamic fundamental frequency.

[0035] The empirical value of 0.5 Hz is used as a hyperparameter, but it can also be adjusted by the implementer according to the actual knitting speed fluctuation range and the system's anti-aliasing requirements.

[0036] The constraints are: , The minimum effective frequency (0.05 Hz for example).

[0037] The above-mentioned frequency band selection is achieved by using well-known finite impulse response bandpass filters or infinite impulse response bandpass filters. The design and implementation of the filters are well-known technical contents and will not be described in detail here.

[0038] Adaptive bandpass filter for continuous tension time series Filtering is performed to obtain the eigenvectors of high-frequency interlaced vibrations. ,Right now: In the formula, For bandpass filtering operation, This is a continuous tension time series, with units of Newtons. The separated high-frequency vibration components reflect the yarn vibration characteristics related to the interlacing cycle.

[0039] The bandpass filter adjusts its position according to the current condition in each control cycle. Recalculate The filter coefficients are updated so that the center frequency of the filter's passband matches the spindle speed changes in real time, thereby reducing spectral leakage caused by speed drift.

[0040] After obtaining the high-frequency vibration components, the low-frequency roll diameter drift trend vector is obtained by removing the separated high-frequency components from the original tension time series, specifically by time-domain subtraction: In the formula, The low-frequency drift component, measured in Newtons, represents the low-frequency components corresponding to the gradual change in yarn diameter consumption and tension.

[0041] The subtraction operation implicitly assumes that bandpass filtering preserves frequency components within the passband well and has minimal impact on low-frequency components in the original signal. Therefore, the difference result can reflect the remaining slowly varying characteristics of the original signal after removing interleaved vibrations.

[0042] During the decomposition process using an adaptive bandpass filter, the filter's inherent phase frequency characteristics will lead to high-frequency components. This introduces a certain group delay, which varies with the filter order and the current center frequency.

[0043] To maintain the phase correspondence between high-frequency components and actuator actions in subsequent processing, this step records the group delay estimate of the adaptive bandpass filter. The unit is seconds, and this estimate can be obtained through the phase frequency response of the filter or simulation calculation.

[0044] The group delay record is only used for time alignment of subsequent feedforward damping compensation commands. It does not change the rule of determining the dynamic fundamental frequency and constructing the bandpass filter based on the real-time spindle speed, nor does it replace the adaptive bandpass filter with a fixed center frequency filter.

[0045] Through the above processing, high-frequency vibration components and low-frequency drift components that match the current spindle operating conditions are obtained in each control cycle, thus completing the dynamic decoupling of the continuous tension signal.

[0046] S3: Calculate the local information entropy deviation and envelope mean of the high-frequency vibration component within the sliding window. When the local information entropy deviation is within the preset allowable range of information entropy deviation, generate a feedforward compensation command based on the envelope mean and the real-time spindle speed; otherwise, limit the increment of the feedforward compensation command. The preset allowable range of information entropy deviation is the upper limit of the preset information entropy deviation.

[0047] In one embodiment, since the high-frequency vibration component not only includes periodic vibrations related to the mechanical interlacing cycle, but may also contain non-periodic disturbances such as local yarn knots, sensor spikes, or short-term impacts, directly generating feedforward compensation with instantaneous amplitude or envelope would amplify these disturbances into erroneous damping commands. Therefore, this step performs criterion analysis on the energy distribution pattern of the high-frequency vibration component within the sliding time window, and decides whether to update the feedforward compensation command and its increment based on the criterion results.

[0048] Before proceeding with the sliding window analysis, the adaptive bandpass filter introduces a group delay estimate during the decomposition process. ,make There is a time lag compared to the actual high-frequency vibration.

[0049] To avoid feedforward compensation and vibration phase misalignment, firstly based on Time alignment of high-frequency vibration components.

[0050] Specifically, Time series forward shift The aligned high-frequency vibration sequence was obtained. .

[0051] The translation operation is achieved by shifting the samples, and the corresponding time offset is equal to the estimated value of the filter group delay. This processing method is a well-known technique and will not be described in detail here.

[0052] Then, a sliding time window is constructed with the current control cycle as the center, and the window length is taken as an empirical value. Each sampling point corresponds to a duration of approximately 2 seconds (control period 20ms). This duration covers several interleaved vibration cycles to ensure statistical stability.

[0053] The window contains recent Aligned high-frequency vibration component samples .

[0054] In each control cycle, the window slides forward as the latest sample arrives.

[0055] After the window is constructed, samples marked as abnormal are removed from the window based on their quality labels. The remaining samples constitute the valid sample set, the size of which is denoted as . .

[0056] like Total number of points in the window If the ratio is lower than the preset effective ratio threshold (0.8 for example, dimensionless), it is determined that the quality of the current window samples is insufficient to support reliable feedforward criteria, and the process directly jumps to the process of limiting the feedforward compensation increment, and the reference information entropy is not updated in this round.

[0057] If the window sample is valid, calculate the envelope mean of the high-frequency vibration components within the sliding window.

[0058] The envelope mean is expressed as the arithmetic mean of the absolute amplitudes of the valid samples: In the formula, The local envelope mean, in Newtons, represents the average energy level of the high-frequency vibration components within the current window. For the first in the window 100 valid aligned sample values, in Newtons.

[0059] Meanwhile, to measure the uniformity of the distribution of high-frequency amplitudes within the window, the absolute amplitude of the effective samples is... Divided into equal intervals according to their numerical range A range (exemplary) ), and calculate the proportion of the sample size in each interval to the total valid sample size. This constitutes an amplitude distribution.

[0060] Calculate the local information entropy on this distribution: In the formula, The local information entropy of the current window is dimensionless and reflects the dispersion of high-frequency amplitudes: when the amplitudes within the window are concentrated in a few intervals, A lower amplitude corresponds to stable periodic oscillations; when the amplitude is dispersed across multiple intervals, A higher value corresponds to an unstable disturbance.

[0061] For the first The normalized proportion of each amplitude interval, dimensionless; if there are no samples in a certain interval, then Corresponding The item is set to 0.

[0062] Reference Information Entropy The baseline entropy value is used to characterize the amplitude distribution of historically stable interlaced windows.

[0063] When the system is running for the first time or when no valid stable window has been established before, the current window will be... As initial However, it is not used to generate feedforward compensation at this time, but only to establish a reference line.

[0064] In subsequent operation, Update smoothly only when the window meets preset stability conditions.

[0065] Stability conditions include: the current window samples are valid, the local information entropy deviation is within the allowable range, and the high-frequency envelope changes within the window do not manifest as isolated shocks.

[0066] The method for determining isolated impacts is as follows: if within the window... The maximum value exceeds the envelope mean. If the preset multiple (e.g., 3, dimensionless) is used, it is considered that a significant local impact has occurred, and the stability condition is not met.

[0067] When all the above stability conditions are met, the reference information entropy is updated smoothly using first-order recursion: In the formula, The smoothing coefficient is taken as an empirical value of 0.9. It is dimensionless and can also be adjusted by the implementer according to the rate of change of the working conditions.

[0068] After the update, It can slowly follow the changes in the interlaced vibration pattern under normal operating conditions, while avoiding jumps caused by a single disturbance.

[0069] get After, and the present Compare and calculate the local information entropy deviation: In the formula, The local information entropy deviation is dimensionless. The smaller its value, the closer the amplitude distribution of the current window is to the historical stable state.

[0070] Preset an allowable deviation threshold The empirical value is 0.2, which is dimensionless and is a hyperparameter that can also be adjusted by the implementer.

[0071] when If the sample validity and rotational speed validity of the current window are both passed, it is determined that the current window mainly corresponds to the trackable periodic interlacing vibration, and new feedforward compensation commands can be generated based on the window.

[0072] Otherwise, if the current window is determined to contain unstable disturbances or abnormal vibration patterns, the increment of the feedforward compensation command is limited to avoid amplifying the disturbance.

[0073] When new compensation commands are allowed to be generated, the feedforward damping compensation factor is first calculated using the envelope mean and the real-time spindle speed: In the formula, is the feedforward damping compensation factor, which is dimensionless and represents the required damping compensation coefficient calculated based on the current vibration energy and rotational speed. This is the mean of the local envelope, expressed in Newtons. For the damping mapping gain, an empirical value of 0.05 is used. , which are preset parameters, convert the force amplitude value into a dimensionless compensation base value; The real-time spindle speed, measured in revolutions per minute; The speed sensitivity index is taken as an empirical value of 0.002 min / r, which is a preset parameter used to reflect the nonlinear effect of speed changes on the feedforward compensation strength.

[0074] and All of these can be determined by equipment calibration or historical data, and implementers can also adjust them according to the actual response characteristics of the damping actuator.

[0075] Then, the compensation factor is converted into a feedforward compensation instruction that the actuator can accept through a linear mapping: In the formula, It is a feedforward compensation command, a dimensionless control quantity; The feedforward mapping gain is taken as an empirical value of 1, which is dimensionless and is a preset parameter. This is the feedforward mapping bias, taken as an empirical value of 0, dimensionless, used to compensate for the static deviation of the actuator.

[0076] and The settings can be identified and adjusted through the actuator calibration curve or the system.

[0077] when If any of the following limitations are met, such as insufficient valid samples in the window, abnormal spindle speed, or isolated impact, the control unit will not update the current window. Transformed into a new Instead, it limits the increment of feedforward compensation instructions.

[0078] One of the limiting methods is to keep the feedforward compensation command from the previous stable cycle unchanged, i.e. This makes the feedforward increment close to zero; the second method is to reduce the increment according to a preset attenuation ratio. If the instruction increment that should be calculated based on the current window is... The actual increment is then taken as ,in The attenuation coefficient is taken as an empirical value of 0.2, which is dimensionless and can also be selected by the implementer.

[0079] In this embodiment, the method of maintaining the previous stable instruction is preferred to avoid any residual disturbance entering the actuator.

[0080] Based on the above criteria, the current cycle selects to output an updated feedforward compensation command or a restricted feedforward command according to the vibration mode, so that the feedforward compensation effect only acts on the stable interlaced vibration window, reducing the amplification of unstable disturbances.

[0081] S4: Using a preset calculation cycle as the unit, extract the representative value of the low-frequency drift component in the current calculation cycle, calculate the drift gradient based on the representative values ​​of the current calculation cycle and the previous calculation cycle, accumulate the drift gradient to obtain the drift accumulation amount, and determine the continuity of the drift direction. When the drift accumulation amount exceeds the preset drift threshold and the drift direction is continuous, correct the preset feedback proportional gain; otherwise, maintain the feedback proportional gain before correction.

[0082] In one embodiment, during the operation of the packing braiding machine, the yarn roll diameter decreases slowly as the braiding process is consumed. This process is characterized by a continuous, directionally biased, and gradual change in the low-frequency drift component.

[0083] However, random friction in the yarn path, local knots, or short-term sensor vibrations can also cause problems. This results in non-continuous amplitude fluctuations.

[0084] If directly using two adjacent control cycles The difference is used as a basis for adjusting the feedback proportional gain. If an update is performed, the aforementioned short-term fluctuations will trigger... The disordered jumps cause additional oscillations in the feedback control command under normal roll diameter consumption, and exacerbate the risk of closed-loop oscillations near the actuator boundary.

[0085] Therefore, this step is for The amplitude and direction of change over multiple consecutive control periods are statistically analyzed to distinguish between persistent drift and random fluctuations, and the decision to make corrections is based on the statistical results. .

[0086] First, extract the sliding window representative value of the low-frequency drift component within the current control cycle.

[0087] A control cycle contains several discrete... Sampling points: This step uses all points marked as valid (non-abnormal) within this period. The arithmetic mean of the samples is used as a low-frequency representative value for the current period.

[0088] The marking rules for abnormal samples are as follows: During the tension data acquisition process, samples are marked based on whether they have communication gaps, obvious isolated spikes, or whether they exceed the effective range of the sensor. Samples marked as abnormal are not included in the representative value calculation. The criteria for judging obvious isolated spikes are: if the difference between the amplitude of a sampling point and the average amplitude of the adjacent sampling points exceeds a preset spike amplitude threshold (exemplarily taken as 3 times the amplitude standard deviation within the sliding window), and the duration of the abnormal amplitude does not exceed a preset spike duration threshold (exemplarily taken as 2 sampling periods), then the sampling point is marked as an obvious isolated spike.

[0089] The current period representative value is calculated as follows: In the formula, This is the low-frequency representative value of the current control cycle, in Newtons, reflecting the average tension level of the low-frequency drift component within this cycle; For the current period, the first Low-frequency drift sample values ​​that were not marked as abnormal, in Newtons; This represents the number of valid samples used in the calculation within the current period.

[0090] If the number of valid samples in the current period is lower than the preset minimum number of samples (an empirical value of 3, dimensionless, and a hyperparameter), then the low-frequency representative value of the current period is considered invalid. It is not updated, and the drift gradient and drift accumulation are not calculated based on the current period; the feedback proportional gain is not applied. The original value remains unchanged.

[0091] Low-frequency representative value of the previous cycle The same extraction method is used, and when the representative value of the previous period is invalid, the drift gradient is not calculated.

[0092] After obtaining the representative low-frequency values ​​of two adjacent effective periods, calculate the low-frequency drift gradient of the adjacent periods: In the formula, The drift gradient is expressed in Newtons. A positive value indicates that the low-frequency tension in the current cycle is increasing relative to the previous cycle, while a negative value indicates that it is decreasing. The absolute value reflects the magnitude of the change in low-frequency tension between the two control cycles.

[0093] The sign and magnitude are the basic units for judging the cumulative drift and the continuity of direction.

[0094] To capture the persistent, slowly varying characteristics caused by convolutional diameter consumption, this step accumulates the absolute change in the drift gradient within a sliding observation interval that includes multiple consecutive control cycles.

[0095] Set the observation interval length M, take an empirical value of 5, which is dimensionless and is a hyperparameter. It can also be adjusted by the implementer according to the time scale of the roll diameter consumption process.

[0096] The drift gradients calculated for the most recent M effective control cycles (i.e., representative values ​​for each cycle are effective). Calculate the cumulative drift: In the formula, The cumulative drift is expressed in Newtons and represents the overall intensity of the low-frequency drift component change over M consecutive periods.

[0097] This cumulative method does not distinguish between upward and downward directions, but only measures the total amount of cumulative change, thus differentiating between continuous drift in one direction and multiple periods and dispersed changes with frequent fluctuations in both directions.

[0098] If any of the most recent M periods contain invalid representative values, then those invalid periods are not included in the accumulation. Simultaneously, the M-window is traced back to supplement the valid periods until M valid periods are reached. If the number of valid cycles that can be traced is less than M, the correction judgment of the feedback proportional gain will not be triggered for the time being, and the previous stable gain will be maintained directly.

[0099] In addition to the cumulative change, it is also necessary to confirm whether the direction of low-frequency drift is stable and consistent, so as to avoid misjudging recent alternating up and down fluctuations as continuous drift.

[0100] Within M effective periods, for each effective period... Extract its symbols : In the formula, For a sign function, when hour Take the positive, when hour Take the negative; if The absolute value is less than the preset near-zero threshold. (Taking an empirical value of 0.01N as the preset parameter), then the period... Assuming no significant direction, the corresponding Directional statistics are not included.

[0101] Count the number of positive symbols within M effective periods. and the number of negative signs And calculate the total number of valid symbols. .

[0102] like If the value is 0, it is considered that the direction cannot be determined, and the previous stable gain is maintained directly.

[0103] when When, the number of main direction symbols is defined as And calculate the proportion of the main direction. .

[0104] At the same time, check two adjacent valid symbols in chronological order. and Did positive and negative reversal occur? Count the number of reversals. .

[0105] Preset direction ratio threshold Take an empirical value of 0.8, dimensionless, and set an upper limit on the number of allowed reversals. The empirical value is 1, which is dimensionless. Both are hyperparameters and can be adjusted by the implementer according to the strictness of the requirements for directional stability.

[0106] If and only if and If the low-frequency drift direction is continuous within M consecutive cycles, it is determined that the drift direction is discontinuous; otherwise, it is determined that the drift direction is discontinuous.

[0107] Continuous direction means that changes in low-frequency tension have a consistent trend, which can establish a causal correspondence with deterministic, slowly changing factors such as continuous consumption of yarn diameter and slow yarn elongation.

[0108] Furthermore, a preset drift safety threshold is established. The preset parameter is 0.5N, which is an empirical value. It can also be adjusted by the implementer according to the actual tension control accuracy requirements.

[0109] When the drift accumulation Exceed Furthermore, when the above-mentioned directional continuity determination is valid, it indicates that the low-frequency tension has formed a continuous drift with sufficient amplitude and stable direction due to factors such as the consumption of the roll diameter.

[0110] At this point, the system corrects the proportional gain of the feedback controller to obtain control sensitivity that matches the current drift level.

[0111] The correction method is as follows: In the formula, The corrected proportional gain coefficient is dimensionless. The basic proportional gain is taken as an empirical value of 2, which is dimensionless and is a preset parameter, tuned according to the open-loop steady-state characteristics of the braiding machine tension; The trend compensation convergence factor is set to an empirical value of 0.1. , which is a preset parameter that controls the adjustment weight of the drift gradient with respect to the proportional gain. It can also be adjusted by the implementer according to the convergence speed requirements and overshoot suppression requirements of the control system. To represent the gradient value of the current drift trend, the current effective period can be used. Alternatively, the arithmetic mean of all effective gradients in the same direction within M effective periods can be used (when the mean is used, the influence of random fluctuations in the gradient of a single period on the gain correction can be further suppressed).

[0112] The unit is Newton. A dimensionless gain increment is generated and superimposed on the base proportional gain to obtain an updated proportional gain that matches the current drift intensity and direction.

[0113] when If the drift direction is discontinuous, or if there are invalid representative values ​​within the most recent M effective cycles, resulting in the inability to construct a complete drift accumulation and direction determination, the proportional gain is not corrected; instead, the feedback controller maintains the previous stable gain. Unchanged, that is .

[0114] The previous stable gain can be the most recently successfully corrected gain in the historical effective period. It can also be the base proportional gain. (When no effective correction has ever occurred).

[0115] This holding mechanism ensures that scattered drift information caused by short-term disturbances, instantaneous sensor jumps, or insufficient effective samples will not be transmitted to the feedback gain parameters, thereby preventing the closed-loop system from becoming unstable due to frequent parameter changes.

[0116] After this step is completed, the proportional gain coefficient of the current control cycle It has been set to [value] based on the drift accumulation and direction continuity decision results. Or maintain the original value.

[0117] This gain value is then written to the feedback controller, and in the current control cycle and subsequent control cycles, the feedback controller adjusts it based on the target tension. and current low-frequency representative value Generate feedback control commands.

[0118] when When a valid sample cannot be obtained due to insufficient data, the feedback controller uses the previous valid representative value or a preset safety tension value to participate in the feedback calculation, so as to avoid sudden changes in instructions due to missing input.

[0119] S5: The feedforward compensation command is superimposed with the feedback control command generated based on the feedback proportional gain and normalized tension error, and nonlinear smoothing is performed according to the upper and lower limits of the actuator's saturated output to obtain the final tension control command; wherein, the nonlinear smoothing is as follows: when the superimposed command value is between the upper and lower limits of the saturated output, it remains unchanged; when it exceeds the limits, a continuously differentiable saturation function is used to compress the command value to between the upper and lower limits of the saturated output.

[0120] In one embodiment, the normalized tension error is a dimensionless quantity obtained by dividing the difference between the target tension and the current calculation period representative value of the low-frequency drift component by a preset normalized reference tension.

[0121] The saturation function is a monotonically increasing exponential asymptotic function. When the superimposed instruction value is much smaller than the upper limit of the saturation output, it is approximately linearly transmitted. When the superimposed instruction value is close to or exceeds the upper limit of the saturation output, it asymptotically converges to the upper limit of the saturation output.

[0122] The feedforward compensation command obtained from step S3 and the feedback control command obtained from step S4 This step involves synthesis and the application of boundary constraints.

[0123] Among them, the feedback control command is based on the current proportional gain. Generated from the core, It has been set to [value] in step S4 based on the low-frequency drift accumulation and direction continuity decision results. Or maintain the previous stable gain.

[0124] The specific process for generating feedback control commands is as follows: Take the target tension. The unit is Newtons, and this value is given by the tension setting of the packing braiding machine; it is taken as the low-frequency representative value of the current control cycle. The unit is Newton, derived from the output of step S4; calculate the normalized tension error. ,in The normalized reference tension is set to an empirical value of 1N, which is a preset parameter. Its function is to convert the original force error into a dimensionless relative value so that the subsequent control dimensions are unified.

[0125] Will Compared with the current proportional gain Multiplying them together yields the feedback control command: In the formula, The feedback control command is dimensionless and reflects the strength of the control action required based on the low-frequency drift error. The current proportional gain is dimensionless, and its value is determined in step S4 by... And drift correction terms are determined; The normalized tension error is dimensionless.

[0126] When due to insufficient valid samples When unavailable, The normalized error from the previous effective period or a preset safety value (set to 0, dimensionless) is used to maintain... The continuous and stable flow.

[0127] get Then, it is combined with the feedforward compensation command. By superimposing these commands, the original control instructions are obtained: In the formula, The original control commands after superposition are dimensionless.

[0128] because and All of them are dimensionless quantities and can be directly added algebraically.

[0129] The damping actuator of the packing braiding machine has a physical saturation boundary, and its maximum allowable output amplitude corresponds to the upper limit. .

[0130] In this embodiment, a dimensionless normalized value is used, taking the empirical value of 1. It is dimensionless and is a preset parameter, corresponding to the maximum control equivalent of the actuator. Its value is determined based on the actuator calibration data.

[0131] If left untreated, directly Restricted by the truncation method Within the interval, when If the boundary is slightly exceeded, the actuator command will undergo a step jump, causing an impact on the actuator and a sudden change in yarn tension.

[0132] Therefore, this step is for Apply a nonlinear smoothing mapping to continuously and asymptotically compress the original instructions to within the saturation boundary.

[0133] Nonlinear smoothing uses the following mapping relationship: In the formula, The final tension control command is dimensionless and is the actual control signal sent to the actuator. The upper limit of the actuator saturation boundary is dimensionless; The original control commands after superposition are dimensionless; It is a natural exponential function.

[0134] This mapping function has a monotonically increasing property, when much smaller hour, near , making Control commands are transmitted approximately linearly; when Increase and approach or exceed When the exponential term approaches 0, Approaching step by step However, it will not exceed the limit, achieving smooth convergence to the upper limit of saturation and avoiding execution abrupt changes caused by truncation limiting.

[0135] like A negative value indicates that the control system tends to reduce or even reverse the output after superposition. For actuators that can only output positive damping, there is no reverse physical meaning, so it is directly... Set to 0 to prevent invalid negative instructions from entering the executor.

[0136] This zero-cut operation is limited to the negative half-axis and does not disrupt the smoothness of the positive region.

[0137] In actual system operation, if the actuator saturation boundary parameters If calibration data is missing or sensor malfunction prevents its acquisition, a preset safety limit value will be used. Take the empirical value 1, which is dimensionless, and replace it. The system participates in the above nonlinear smoothing calculation, and at the same time sets a "boundary untrusted" flag. During the existence of this flag, the amplification of the feedforward compensation factor in step S3 and the expansion and update of the feedback proportional gain in step S4 are suspended to prevent the control amplitude from growing excessively when the boundary is unknown.

[0138] After the above processing, the final tension control command is obtained in the current control cycle. The command is then sent to the damping actuator via the industrial fieldbus or analog output channel to complete the tension adjustment action.

[0139] because It is compatible with actuator physical constraints, and the actuator will not be impacted by sudden changes in commands during the response process. The yarn tension can still maintain a smooth transition when approaching the control boundary.

[0140] The tension adaptive control system for a packing braiding machine includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the tension adaptive control method for a packing braiding machine according to the first aspect of the present invention.

[0141] The tension adaptive control system of the packing braiding machine also includes other components well known to those skilled in the art, such as communication buses and communication interfaces. Their settings and functions are known in the art and will not be described in detail here.

[0142] It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept, and these all fall within the scope of protection of this invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A tension adaptive control method for a packing braiding machine, characterized in that, include: Acquire the real-time spindle speed and continuous tension signal, and obtain the upper limit and lower limit of the actuator's saturated output; Based on the real-time spindle speed, the interlacing dynamic fundamental frequency is determined. An adaptive bandpass filter is constructed with the interlacing dynamic fundamental frequency as the center frequency to decompose the continuous tension signal and obtain the high-frequency vibration component and the low-frequency drift component. Within a sliding window, the local information entropy deviation and envelope mean of the high-frequency vibration component are calculated. When the local information entropy deviation is within the preset allowable range, a feedforward compensation command is generated based on the envelope mean and the real-time spindle speed. Otherwise, the increment of the feedforward compensation command is limited. The preset allowable range of information entropy deviation is that the local information entropy deviation does not exceed the preset upper limit of information entropy deviation. Using a preset calculation cycle as the unit, extract the representative value of the low-frequency drift component in the current calculation cycle, calculate the drift gradient based on the representative values ​​of the current calculation cycle and the previous calculation cycle, accumulate the drift gradient to obtain the drift accumulation amount, and determine the continuity of the drift direction. When the drift accumulation amount exceeds the preset drift threshold and the drift direction is continuous, the preset feedback proportional gain is corrected; otherwise, the feedback proportional gain before correction is maintained. The feedforward compensation command is superimposed with the feedback control command generated based on the feedback proportional gain and normalized tension error, and nonlinear smoothing is performed according to the upper and lower limits of the actuator's saturated output to obtain the final tension control command. The nonlinear smoothing process is as follows: when the superimposed command value is between the upper and lower limits of the saturated output, it remains unchanged; when it exceeds the limits, a continuously differentiable saturation function is used to compress the command value to between the upper and lower limits of the saturated output.

2. The tension adaptive control method for a packing braiding machine according to claim 1, characterized in that, The determination of the interleaved dynamic fundamental frequency includes: The interlacing dynamic fundamental frequency is calculated based on the preset transmission ratio coefficient and the real-time spindle speed. The transmission ratio coefficient is the mechanical conversion ratio from the main shaft speed to the interlacing frequency.

3. The tension adaptive control method for a packing braiding machine according to claim 1, characterized in that, The adaptive bandpass filter constructed using the interleaved dynamic fundamental frequency as the center frequency includes: Preset frequency band isolation tolerance width; An adaptive bandpass filter is constructed by using the difference between the interleaved dynamic fundamental frequency and the bandwidth of the band isolation tolerance as the lower cutoff frequency and the sum of the interleaved dynamic fundamental frequency and the bandwidth of the band isolation tolerance as the upper cutoff frequency. Obtain the group delay of the adaptive bandpass filter; Before generating the feedforward compensation command, phase compensation is performed on the high-frequency vibration components based on the group delay.

4. The tension adaptive control method for a packing braiding machine according to claim 1, characterized in that, The calculation of the local information entropy deviation and envelope mean of the high-frequency vibration components includes: Calculate the mean of the absolute amplitudes of the high-frequency vibration components within the sliding window, and use it as the envelope mean. Within the sliding window, the amplitude of the high-frequency vibration component is divided into equal-width bins according to the preset number of bins. The sample proportion in each bin is counted as the amplitude distribution. Based on the amplitude distribution, the local information entropy is calculated according to the Shannon information entropy formula. Obtain the reference information entropy; The absolute value of the difference between the local information entropy and the reference information entropy is taken as the local information entropy deviation.

5. The tension adaptive control method for a packing braiding machine according to claim 4, characterized in that, The acquisition of reference information entropy includes: Obtain the entropy of existing reference information in the current period; Calculate the local information entropy deviation of the current period based on the local information entropy of the current period and the existing reference information entropy of the current period. When the sliding window meets the preset stability conditions, the local information entropy of the current period is updated to the new reference information entropy. Otherwise, the existing reference information entropy in the current cycle remains unchanged; The preset stability conditions include: the real-time spindle speed within the sliding window is valid, the continuous tension signal does not have missing data or exceed the preset signal validity range, the local information entropy deviation of the current cycle is within the preset information entropy deviation allowable range, and the envelope change of the high-frequency vibration component does not manifest as an isolated impact; wherein, an isolated impact is defined as: the change in the mean envelope value within the sliding window relative to the mean envelope value of the previous sliding window exceeds the product of the preset impact judgment multiple and the mean envelope value of the previous sliding window, and the change falls back to within the preset fallback ratio of the mean envelope value of the previous sliding window in the next sliding window.

6. The tension adaptive control method for a packing braiding machine according to claim 1, characterized in that, The generated feedforward compensation instruction includes: Based on the envelope mean, real-time spindle speed, preset damping mapping gain, and speed sensitivity index, a feedforward damping compensation factor is generated by multiplying the damping mapping gain by the product of the envelope mean and the speed sensitivity index of the real-time spindle speed raised to the power of the speed sensitivity index. The speed sensitivity index is used to characterize the degree of nonlinear influence of spindle speed variation on the feedforward damping compensation strength. The feedforward damping compensation factor is linearly mapped using a preset feedforward mapping gain and a preset bias to obtain the feedforward compensation command; wherein, the linear mapping is the product of the feedforward mapping gain and the feedforward damping compensation factor plus the preset bias.

7. The tension adaptive control method for a packing braiding machine according to claim 1, characterized in that, The increment of the restricted feedforward compensation command includes: When the local information entropy deviation exceeds the preset allowable range of information entropy deviation, the increment of the feedforward compensation instruction of the previous calculation cycle is used as the increment of the feedforward compensation instruction of the current calculation cycle, or the increment of the feedforward compensation instruction of the current calculation cycle is reduced according to the preset attenuation ratio.

8. The tension adaptive control method for a packing braiding machine according to claim 1, characterized in that, The determination of drift direction continuity includes: The drift gradient symbols are counted within a preset number of consecutive calculation cycles. When the proportion of the number of symbols in the main direction to the number of effective symbols reaches the preset proportion and the number of symbol reversals does not exceed the preset number of reversals, the drift direction is determined to be continuous.

9. The tension adaptive control method for a packing braiding machine according to claim 1, characterized in that, The correction of the preset feedback ratio gain includes: The updated feedback proportional gain is calculated based on the feedback proportional gain, the preset trend compensation convergence factor, and the drift gradient. The updated feedback proportional gain is the original feedback proportional gain plus the product of the trend compensation convergence factor and the drift gradient.

10. A tension adaptive control system for a packing braiding machine, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement the tension adaptive control method for a packing braiding machine according to any one of claims 1-9.

Images