Intelligent quality evaluation method and system for spinning products

Abstract

The invention relates to the technical field of textile engineering, in particular to an intelligent quality evaluation method for spinning products, which comprises the following steps: S1, collecting and preprocessing a multi-channel capacitance signal to obtain a standardized capacitance signal; and S2, constructing a sum-difference capacitance signal based on the standardized capacitance signal. In order to solve the technical problem of yarn evenness unevenness measurement distortion caused by yarn transverse swing, tension fluctuation and speed change in cone yarn on-line detection of an existing capacitive yarn evenness instrument, multi-channel capacitance signals are collected, linear density and position information are decoupled through sum-difference combination and linear inverse operation, and yarn evenness unevenness measurement distortion is obtained. And position nonlinear compensation, tension effect compensation and resampling mapping from a time domain to a length domain are carried out in sequence, and finally the accurate evenness unevenness is calculated. The interference introduced by a measurement system and a process working condition is effectively stripped, so that the evaluation result truly reflects the quality fluctuation of the yarn, and the accuracy and the reliability of online quality detection are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of textile engineering, in particular to an intelligent quality evaluation method and system for spun yarn products. BACKGROUND

[0002] As an intermediate product in the textile production process, the stability of yarn quality indicators such as evenness, defect number, hairiness index, and strength is crucial. For a long time, the quality evaluation of spun yarn products has mainly relied on manual detection and traditional mechanical electronic detection equipment. The manual detection method usually involves quality inspectors observing the yarn surface or fabric sample under specific lighting conditions with the naked eye to determine whether there are defects such as impurities, broken ends, and color differences. However, this detection method has many limitations: first, it is highly subjective, and differences in experience, vision, and fatigue levels of different inspectors can lead to different results, making it difficult to establish a unified quantitative standard; second, it is inefficient and cannot meet the high-flow detection needs of modern high-speed spinning production lines; third, manual detection cannot capture subtle and intermittent quality fluctuations, leading to missed or false detections. To improve the objectivity and efficiency of detection, capacitive or photoelectric sensors are introduced in existing technologies for offline or online sampling detection.

[0003] On a ring spinning production line, in order to monitor the quality of the wound yarn product online, a capacitive evenness tester is usually configured in the winding process to continuously detect the evenness of the high-speed running yarn, and the yarn evenness irregularity is used as the core quality evaluation parameter. This parameter reflects the fluctuation of the instantaneous linear density of the yarn relative to the average linear density from the capacitive signal, and is used to evaluate the uniformity of the yarn within a certain measurement length range. It is one of the most basic and critical physical quantities in spinning process control, quality grading, and intelligent evaluation models.

[0004] However, under existing technical conditions, capacitive evenness testers are mainly used for online detection of wound yarn or large packages. During the winding process, due to factors such as the reciprocating horizontal movement of the yarn guide, the geometric changes between the layers of the package, and the differences in winding tension at different positions on the bobbin, the yarn cannot always remain strictly centered between the electrodes, but instead makes periodic lateral oscillations within the measurement port. Existing evenness testers generally equate the change in capacitance directly to the change in linear density for evenness irregularity calculation, without distinguishing between the real linear density fluctuation of the yarn and the false signal introduced by the position shift coupling of the electric field. This results in the inevitable mixing of false fluctuations caused by position shifts in the measured evenness signal under complex winding conditions.

[0005] Therefore, an intelligent quality evaluation method and system for spun yarn products are proposed to solve the above-mentioned problems. SUMMARY

[0006] Technical problems solved

[0007] In view of the above-mentioned defects of the prior art, the present application provides an intelligent quality evaluation method and system for a spun yarn product, which can effectively solve the problem of online detection of sliver unevenness distortion in the prior art, and ultimately achieve the purpose of accurately evaluating the inherent quality of the yarn.

[0008] Technical scheme

[0009] To achieve the above object, the present application is implemented by the following technical scheme:

[0010] The present application provides an intelligent quality evaluation method and system for a spun yarn product, comprising the following steps:

[0011] S1: collecting multi-channel capacitance signals and performing preprocessing to obtain standardized capacitance signals;

[0012] S2: constructing sum and difference capacitance signals based on the standardized capacitance signals;

[0013] S3: establishing a linear relationship between the sum and difference capacitance signals and the yarn linear density and transverse position;

[0014] S4: based on the linear relationship, obtaining an instantaneous linear density sequence and a transverse position sequence through inverse operation;

[0015] S5: performing position correction on the instantaneous linear density sequence to obtain a compensated linear density sequence;

[0016] S6: mapping the compensated linear density sequence from the time domain to the length domain to obtain a length-equidistant linear density sequence;

[0017] S7: calculating the average linear density based on the length-equidistant linear density sequence;

[0018] S8: generating a zero-mean fluctuation sequence based on the average linear density;

[0019] S9: calculating the linear density variance and standard deviation based on the zero-mean fluctuation sequence;

[0020] S10: calculating the sliver unevenness based on the standard deviation and the average linear density.

[0021] Further, the preprocessing in S1 comprises:

[0022] calculating the standardized capacitance signal ; in the formula, the standardized capacitance signal of the i-th channel at time t; is the original capacitance value measured by the i-th channel at time t; is the reference capacitance of the i-th channel when no yarn passes through; is the calibration gain factor for the i-th channel; i is the index number.

[0023] Furthermore, the sum and difference capacitance signals mentioned in S2 include sum signals. Sum and difference signals ,in: In the formula and These are the standardized capacitor signals for the first and second channels, respectively.

[0024] Furthermore, the linear relationship described in S3 is expressed as:

[0025] In the formula, Let be the linear density of the yarn passing through the electrode cross-section at time t; Let t be the transverse position of the yarn within the electrode cross-section; For constant terms; The sensitivity coefficient of the signal to line density; Let be the residual sensitivity coefficient of the signal to lateral position; The sensitivity coefficient of the difference signal to the line density; The residual sensitivity coefficient of the difference signal to the lateral position.

[0026] Furthermore, the inverse operation described in S4 includes:

[0027] The instantaneous linear density sequence and the lateral position sequence are obtained by inverse operation of the linear mapping matrix, wherein... M is the calibrated linear mapping matrix.

[0028] Furthermore, the position correction described in S5 includes:

[0029] The instantaneous linear density sequence is corrected using a position correction function to obtain the compensated linear density sequence. ,in ; For the reference gain term, For linear position offset terms, These are quadratic nonlinear terms; all were obtained through calibration using standard yarns.

[0030] Furthermore, S5's position correction of the instantaneous linear density sequence also includes tension compensation, which includes:

[0031] S501: Acquires the tension output signal through a tension sensor and calibrates to obtain the true tension sequence, i.e.: In the formula, For true tension; This is the slope calibration coefficient for the linear relationship between tension and sensor output; This is the intercept calibration coefficient in the linear relationship between tension and sensor output;

[0032] S502: Discretely sample the real tension sequence to obtain the instantaneous tension sequence, i.e.: ;in, The instantaneous tension value at the nth sampling time; For a moment The corresponding actual tension; For at any time The tension output signal;

[0033] S503: Smooth the instantaneous tension sequence to obtain a tension estimate sequence, i.e.: In the formula, Let be the tension estimate at the nth sampling time; Q is the window length of the sliding process; q is the index number of the window. Let be the instantaneous tension value at the nq-th sampling time;

[0034] S504: Calculate the yarn elongation ratio based on the tension estimation value sequence, i.e.: In the formula, This represents the actual length of the yarn segment under the current tension estimate. This is the actual length of the yarn segment under the reference tension; For the strain, i.e.: ;in For additional elongation; As the reference tension; The equivalent compliance within a given tension range;

[0035] S505: Calculate the apparent linear density based on the elongation ratio, i.e.: In the formula, The apparent linear density of the yarn of the current length under tension T; The baseline linear density under reference tension;

[0036] S506: Based on the apparent linear density and reference tension, calculate the true linear density sequence, i.e.: In the formula, This is the actual linear density under the reference tension, derived from the apparent linear density under the current tension.

[0037] Furthermore, S6 further includes mapping the compensated linear density sequence from the time domain to the length domain:

[0038] S601: Collect the instantaneous angular velocity sequence of the winding roller;

[0039] S602: Based on the instantaneous angular velocity sequence and the winding radius recursive model, update the winding radius sequence, i.e.: In the formula, The winding radius at the nth sampling time; The previous sampling time The winding radius; This is the structural proportionality coefficient; The previous sampling time linear velocity;

[0040] S603: Based on the winding radius sequence and instantaneous angular velocity sequence, calculate the original linear velocity sequence, i.e.: In the formula, Let n be the linear velocity at the nth sampling time. The angular velocity at the nth sampling time;

[0041] S604: Smooth the original linear velocity sequence to obtain an effective linear velocity sequence, i.e.: In the formula, M is the effective linear velocity after smoothing; M is the sliding window length. This represents the original linear velocity value at the m-th sampling point counting backwards from the current moment;

[0042] S605: Calculate the cumulative passing length sequence based on the effective linear velocity sequence; that is: In the formula, This is the initial length position; The cumulative length of the yarn relative to the initial point at the nth sampling time; Let N be the linear velocity at the i-th sampling time in the effective linear velocity sequence; N is the total number of samples in the effective linear velocity sequence.

[0043] S606: Based on the aforementioned cumulative length sequence, divide the equidistant length position sequence, that is: In the formula, This indicates the length position of the k-th target length sampling point; is the length sampling interval; k is the index number of the length position; K is the total number of equally spaced length sampling points, i.e. ;in This refers to the total length of the yarn that has passed up to the last time sampling point; Indicates rounding down;

[0044] S607: The compensated linear density sequence is mapped to the equidistant length position sequence through linear interpolation to obtain a linear density sequence with equal length intervals, i.e.:

[0045] In the formula, For the length position Linear density after tension compensation; For at a certain point in time The linear density after tension compensation was measured. For at a certain point in time The linear density is measured and compensated for by tension; finally, a linear density sequence with equal length intervals is obtained.

[0046] Furthermore, generating the zero-mean fluctuation sequence based on the average linear density in S8 also includes removing linear drift, which includes:

[0047] S801: Calculate the sum of the length position sequence ;

[0048] S802: Calculate the sum of the length position square sequence. ;

[0049] S803: Calculate the sum of the linear density sequences with equal length intervals. ;

[0050] S804: Calculate the sum of the product sequence of length position and linear density. ;

[0051] S805: Based on , , and Calculate the slope of the linear trend ;

[0052] S806: Based on , Calculate the linear trend intercept with A. ;

[0053] S807: Based on A and B, detrend the line density sequences with equal length intervals to obtain a detrended line density sequence, i.e.: In the formula, Indicates the position at length Detrending line density at the location;

[0054] S808: Use the detrended linear density sequence as the updated zero-mean fluctuation sequence.

[0055] A smart quality assessment system for spun yarn products, comprising:

[0056] The capacitor signal acquisition module is used to acquire multi-channel capacitor signals and perform preprocessing.

[0057] The sum and difference signal construction module is used to construct sum and difference capacitance signals;

[0058] The linear relationship establishment module is used to establish the linear relationship between the capacitance signal and the physical quantity;

[0059] The inverse operation module is used to calculate the instantaneous linear density sequence and the lateral position sequence;

[0060] The position correction module is used to correct the position of the line density;

[0061] The mapping module is used to map the time domain to the length domain;

[0062] The sequence generation module is used to generate zero-mean fluctuation sequences.

[0063] Beneficial effects

[0064] The technical solution provided by this invention has the following advantages compared with the prior art:

[0065] This invention, by acquiring dual-channel (or multi-channel) capacitance signals and constructing a sum signal sensitive to linear density and a difference signal sensitive to position, and then using a pre-calibrated linear relationship matrix for inverse operation, achieves for the first time the real-time, online calculation of independent yarn instantaneous linear density and lateral position sequences from mixed signals. It overcomes the limitation of traditional single-channel capacitance sensors that treat signal aliasing as a whole, fundamentally solving the industry problem of measurement errors caused by the periodic oscillation of yarn within the measuring slot. By separating the position effect from the linear density signal, subsequent linear density data no longer contains pseudo-fluctuations caused by uneven electric field distribution, significantly improving the accuracy and reliability of online detection. This lays a clean data foundation for all subsequent advanced compensations, allowing indicators such as yarn unevenness to more accurately reflect the quality of the yarn itself.

[0066] After successfully decoupling the lateral position information, the system uses the real-time calculated lateral position sequence to dynamically and nonlinearly compensate for the initially obtained linear density value. This effectively overcomes the inherent nonlinear characteristics of the capacitive sensor's electric field in the edge region. Even if the yarn swings significantly, deviating from the optimal operating range of the linear model, the system can accurately compensate for the resulting changes in measurement gain through this function. This ensures that the measurement system maintains consistent high accuracy in its response to linear density across the entire electrode cross-sectional width, expands the sensor's effective measurement range, and enhances the system's robustness under various complex winding conditions.

[0067] In particular, it was deeply recognized that yarn tension fluctuations cause physical elongation, thus distorting the measured mass per unit length. To address this, tension sensor signals were incorporated into the processing chain. Using a calibrated yarn strain model, the apparent linear density measured under the current tension was converted in real-time to the true linear density under a unified reference tension. This solved the long-neglected problem of tension-induced spurious yarn evenness. By eliminating the false linear density changes introduced by the periodic fluctuations in winding tension, the final quality assessment results reflect only the inherent fluctuations in yarn material structure and thickness, unaffected by changes in process parameters during production. This significantly improves the comparability of quality data from different machines and batches, providing a more reliable basis for process optimization and quality traceability.

[0068] To address the issue that speed fluctuations during high-speed winding cause time sampling to be inconsistent with length sampling, this invention dynamically estimates the instantaneous linear velocity using angular velocity and winding radius. It then establishes a precise mapping from time points to length points by integrating the velocity, and finally generates a linear density sequence with equal length intervals through interpolation algorithms. This ensures that all subsequent statistical and frequency domain analyses are based on real physical length coordinates. This fully conforms to the definition of stripe quality indicators and eliminates the interference of speed fluctuations on unevenness calculations and periodic defect wavelength analysis. The resulting stripe unevenness is a pure quality parameter independent of production speed, making quality assessment results more scientific and standardized, facilitating objective comparisons within the industry. Attached Figure Description

[0069] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0070] Figure 1 This is a schematic diagram of the intelligent quality assessment method in an embodiment of the present invention. Detailed Implementation

[0071] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0072] The present invention will be further described below with reference to embodiments.

[0073] Example:

[0074] Yarn evenness is a typical physical indicator for measuring yarn linear density, which refers to the degree of unevenness in mass or length. Yarn evenness is defined as the degree of variation in instantaneous yarn linear density relative to the average linear density within a given test length, i.e.:

[0075]

[0076] In the formula, U is the yarn evenness rate, which reflects whether the yarn linear density changes evenly with length; the larger the value, the greater the fluctuation in yarn thickness and the worse the quality. For measurement time, the total length of yarn passing through within a certain measurement time is used; This represents the instantaneous linear density signal of the yarn at time t; This represents the average value of the instantaneous linear density signal of the yarn during the measurement time.

[0077] However, in the yarn evenness detection stage of the ring spinning production line, a capacitive evenness meter is used to measure the evenness of the high-speed moving yarn. During the yarn winding process, the yarn is not always strictly centered within the evenness meter's measuring port. Instead, it experiences periodic lateral oscillations due to several influencing factors, including: periodic lateral movement of the forming guide, geometric position changes caused by repeated overlapping of yarn layers, and slight periodic differences in yarn tension between the winding edge and center. This results in a periodic change in the radial position of the yarn within the electric field cross-section of the capacitive sensor over time. Consequently, the measured capacitance signal simultaneously contains both the yarn linear density change signal and the spurious signal caused by the lateral position change of the yarn. These two signals are difficult to distinguish directly in the time domain, leading to a systematic deviation in the yarn evenness rate.

[0078] See appendix Figure 1 This case proposes a smart quality assessment method for spun yarn products, which includes the following steps:

[0079] S1: Acquire multi-channel capacitance signals and preprocess them to obtain standardized capacitance signals that eliminate the influence of environmental and circuit drift, providing a stable and reliable measurement basis for subsequent linear physics calculations; that is: In the formula, The standardized capacitance signal of the i-th channel at time t is the standardized result after zero-point and gain correction of the original capacitance signal, and can be directly compared between different times and different channels. Let be the original capacitance value of the i-th channel measured at time t; This is the reference capacitance of the i-th channel when no yarn passes through; is the calibration gain factor for the i-th channel, reflecting how much the change in capacitance of this channel corresponds to the change in the actual physical quantity, obtained through fitting and calibration of experimental data; i is the index number of the channel.

[0080] In this scheme, dual-channel capacitance signals are acquired and processed, and are respectively denoted as the first channel standardized capacitance signal. Second channel standardized capacitor signal .

[0081] S2: By combining the first-channel standardized capacitance signal and the second-channel standardized capacitance signal through addition and subtraction to construct a sum-difference capacitance signal, signal channels that are more sensitive to line density and lateral position are obtained, respectively. These channels are used to initially separate line density information and position information at the signal layer.

[0082]

[0083] In the formula, It is the sum of the first channel standardized capacitance signal and the second channel standardized capacitance signal at time t, that is, the symmetrical combination result of the dual-channel capacitance signal. It is most sensitive to the total capacitance change determined by the linear density of the yarn, and relatively insensitive to the small left and right swings of the yarn. It is used as the representative signal of the main contribution of linear density. The difference signal between the first channel standardized capacitance signal and the second channel standardized capacitance signal at time t is the antisymmetric combination result of the dual-channel capacitance signals. It is a signal channel that is highly sensitive to the left-right asymmetry of the yarn's lateral position and is used as a representative signal of the main contribution of the lateral position.

[0084] S3: Utilizing the physical laws governing electrode structure and electric field distribution, when the lateral offset of the yarn is relatively small compared to the electrode spacing, a set of linear equations can be used to link the sum and difference capacitance signals with physical quantities. Furthermore, by establishing a linear relationship matrix between the capacitance signal, linear density, and lateral position under calibration conditions, this matrix supports the calculation of physical quantities from electrical signals in the online state; that is:

[0085]

[0086] In the formula, Let be the linear density of the yarn passing through the electrode cross-section at time t; Let t be the transverse position of the yarn within the electrode cross-section; The constant term, i.e., under a given operating condition (e.g., near a certain average line density or average position), and the baseline bias term of the signal, determine the reference level of the signal. The sensitivity coefficient of the signal to linear density reflects how changes in linear density affect the total capacitance. When the yarn remains roughly centered, an increase in linear density will increase the capacitance value of the electrodes on both sides, and thus the signal increases with increasing linear density. The residual sensitivity coefficient of the signal to lateral position is used to quantify how much the signal changes when the yarn is deviated to one side, reflecting the robustness to positional imperfections. The sensitivity coefficient of the difference signal to line density reflects the degree of influence of changes in line density on the difference in capacitance between the left and right sides. The residual sensitivity coefficient of the difference signal to the lateral position quantifies the influence of the lateral offset of the yarn on the difference between the left and right electrodes.

[0087] in , , , and All parameters were determined through fitting experimental data. That is, under a known linear density, the yarn was controlled to be located at different known transverse positions within the electrode cross-section. The sum and difference signals under each experimental condition were recorded and substituted into a linear regression model to solve for the optimal combination of parameters.

[0088] S4: By applying a calibrated linear inverse matrix operation to the real-time sum and difference signals, the instantaneous linear density sequence and lateral position sequence are obtained, which are used to describe the basic physical state of the yarn at the measurement section, i.e.: Where M is the linear mapping matrix between electrical signals and physical quantities, i.e. Used to convert physical quantity vectors Linear mapping to signal vector .

[0089] when When the time is right, the inverse solution can be found. ,Right now:

[0090]

[0091] In the formula, It is the inverse of the linear mapping matrix, i.e. ;in Let be the j-th element in the i-th row of the inverse matrix; thus, the instantaneous linear density sequence and the corresponding transverse position sequence can be obtained.

[0092] S5: By correcting the instantaneous linear density sequence according to its lateral position, a compensated linear density sequence is obtained to eliminate the nonlinear position effect of the electric field. This sequence is used to reflect the true linear density fluctuation of the yarn under ideal centering conditions; that is: In the formula, To compensate for the linear density; This is the position correction function, i.e.: Where y is the input variable of the position correction function; For the reference gain term, when When the yarn is located at the center of the electrode, no additional correction is required, which in turn determines the basic gain level of the reference position; For the linear position offset term, when there is a slight asymmetry in the electrode structure or installation, the effects of the left and right offsets are not completely symmetrical. They will gradually increase or decrease the measurement gain in proportion to y, thus correcting the linear deviation trend caused by the incomplete asymmetry or overall eccentricity of the electrodes. The term is a quadratic nonlinear term, which describes the trend of the error increasing or decreasing in a quadratic manner after deviating from the center. The electric field of the symmetrical electrode is often approximately linear near the center, but the closer to the two edges, the more obvious the changes in field strength and sensitivity, causing the measurement error to rise or fall according to the curvature curve as y increases, which is then used to compensate for the symmetrical nonlinearity of the electrode electric field.

[0093] The position correction function is obtained by selecting a standard yarn with known linear density and controlling the yarn at multiple different lateral positions. The upper part uses a capacitive stripper to collect sufficient data at each position; at each lateral position The linear density obtained from the first-order linear model in S3 is used to calculate its average value; then, the correction factor in the lateral position is calculated, i.e.: ;in, For horizontal position The average estimated linear density at; The actual linear density of the standard yarn is calculated; finally, the discrete points are fitted to obtain a continuous position correction function.

[0094] However, in S5, although the measurement error caused by lateral oscillation is eliminated by using dual-channel capacitance and position correction functions to obtain a compensated linear density sequence, the tension of the yarn is not constant when passing through the measurement area. Yarn segments at different times and positions will be in different stress states. Since the yarn itself is an elongated and compressible medium, the same mass will be stretched under high tension and relatively relaxed under low tension. This means that the mass contained in a unit length will be dynamically redistributed with changes in tension. Consequently, even if the linear density of the yarn body is uniform in the free state, a portion of the linear density curve measured online will appear with pseudo-unevenness introduced by the tension field. That is, the high tension area appears finer and the low tension area appears coarser. The periodically changing tension field along the length direction will be incorrectly recorded as the true linear density fluctuation in the evenness data, thus distorting the unevenness rate and spectral analysis results, making the data from different machines and under different tension conditions incomparable.

[0095] Therefore, by calibrating and smoothing the tension sensor signal, the instantaneous tension sequence at each sampling moment is estimated in real time to reflect the degree of tension experienced by the yarn at the moment of measurement. Then, based on the tension-strain relationship, the apparent linear density observed under the current tension is uniformly converted to an equivalent linear density under a pre-selected reference tension. This is used to eliminate the unit length mass distortion introduced by tension fluctuations, ensuring that the final output yarn evenness truly reflects the inherent unevenness of the yarn structure and is as unaffected as possible by disturbances from process tension conditions. Specifically:

[0096] S501: Acquires tension output signal via tension sensor And by linearly calibrating and changing it, a continuous-time tension signal corresponding to the real physical tension is obtained, that is:

[0097]

[0098] In the formula, True tension directly reflects the magnitude of the force exerted on the yarn at any given moment; It is the slope calibration coefficient for the linear relationship between tension and sensor output, used to stretch or compress the dimensions and proportions of the tension output signal to the order of the true tension. It is the intercept calibration coefficient in the linear relationship between tension and sensor output, used to compensate for zero-point offset, so that the sensor reading under zero tension or a certain reference tension state is correctly translated to the corresponding actual tension value.

[0099] S502: Because capacitance sampling is performed at discrete time points, that is: ;in, This refers to the nth sampling time. is the sampling interval; n is the index number. To establish a one-to-one correspondence between the actual tension and the linear density samples, tension signals need to be taken on the same time axis, thus obtaining the instantaneous tension sequence, i.e.: ;in, The instantaneous tension value at the nth sampling time; For a moment The corresponding actual tension; For at any time The tension output signal.

[0100] S503: Since the tension output signal in actual measurements often contains high-frequency noise, electromechanical vibration, or circuit interference, a moving average smoothing process is applied to the instantaneous tension sequence to obtain a tension estimate sequence with lower noise. This sequence is used to stably reflect the true trend of tension change over time. In the formula, Let be the tension estimate at the nth sampling time; Q is the window length of the sliding process; q is the index number of the window. Let be the instantaneous tension value at the nq-th sampling time.

[0101] S504: Maps the current tension estimate to the yarn elongation ratio to reflect the effect of tension changes on yarn length, i.e.: In the formula, This represents the actual length of the yarn segment under the current tension estimate, reflecting the change in yarn length after tension is applied. It is used to convert strain back from a relative quantity to an absolute length. The actual length of the yarn segment under the reference tension is used as the reference length to measure how much the yarn elongates relative to this reference. The tensile strain represents the proportion of the change in the length of this yarn segment relative to its original length, i.e.: ;in The additional elongation represents the tension from the reference tension. The change in this section of yarn when the current tension estimate is reached; The equivalent compliance within a given tension range, i.e., the strain increment caused by a unit tension increment, is obtained by experimental calibration.

[0102] S505: By relating the true linear density to the elongated length, the apparent linear density observed under the current tension is obtained, reflecting how tension changes the mass per unit length. In the formula, The apparent linear density of the yarn of the current length under tension T; This represents the baseline linear density under the reference tension.

[0103] S506: In S5, the linear density after compensation. Its essence is the apparent linear density under the current tension after position compensation, denoted as: Furthermore, for each time sampling point The true linear density is calculated based on the apparent linear density and tensile strain, resulting in a true linear density sequence. This sequence is used to reconstruct a unified tension baseline from the current tension condition. In the formula, This is the actual linear density under the reference tension, derived from the apparent linear density under the current tension.

[0104] S6: Utilizing instantaneous linear velocity, the real linear density sequence sampled at equal time intervals is mapped and reconstructed into a length-equal-interval sampling sequence, achieving unification from the time domain to the length domain. This is used to accurately reflect yarn evenness fluctuations under real physical length coordinates and calculate reliable yarn evenness indicators; that is: In the formula, Given an online velocity *v* and calculations starting from the starting point, the physical length position of the *n*th sampling point on the yarn is used to map the *n*th sample in time to the *n*th position in length, thus achieving a mapping from time coordinates to length coordinates; *v* is the linear velocity of the yarn as it passes through the measurement section. This allows for the determination of the physical length position. Afterwards, the true linear density sequence can be equivalent to That is, each sampling point is not just the nth time sample, but a sample of length . The linear density sample at the given location forms a discrete linear density distribution curve along the yarn length, which is used to truly reflect the linear density fluctuation of the yarn in space, rather than just an abstract time series signal. This allows all subsequent statistical and frequency domain analyses to be based on physical length rather than time, which is more in line with the physical meaning of the yarn evenness index defined by unit length.

[0105] However, in actual high-speed yarn winding processes, variations in the yarn winding radius over time, slight fluctuations in motor speed, and dynamic response errors in tension control all cause non-negligible fluctuations in the instantaneous linear velocity of the yarn passing through the capacitive measuring port. This results in continuous changes in the actual passing length within the same time interval, causing the evenly spaced sampling on the time axis to appear inconsistent on the length axis. This velocity fluctuation causes what should be uniform length segments to be stretched or compressed non-uniformly in the time series, creating pseudo-stripes unrelated to the actual structure. Furthermore, it causes subsequent statistical and spectral analyses, nominally targeting a specific length, to actually correspond to a mixed range of varying lengths, weakening the effectiveness of position and tension compensation.

[0106] Therefore, the instantaneous linear velocity of the yarn at the measurement point is estimated by collecting the angular velocity of the winding mechanism and combining it with the winding radius. This is used as the basis for subsequently converting the time-equally spaced true linear density sequence into a length-equally spaced signal, thus eliminating the length sampling unevenness caused by velocity fluctuations and ensuring that the yarn evenness analysis truly reflects the intrinsic unevenness of the yarn along its length. Specifically:

[0107] S601: At a fixed sampling interval Below, the instantaneous angular velocity of the winding roller Discrete sampling is performed, and the corresponding angular velocity value is recorded at each sampling time to form a sequence of instantaneous angular velocities.

[0108] S602: The winding radius is updated recursively, accumulating the linear velocity from the previous moment into the winding geometry. This dynamically reflects the change in the yarn bobbin radius as the winding length increases.

[0109]

[0110] In the formula, The winding radius at the nth sampling time represents the value at the current time. The current radius of the yarn wound on the bobbin; The previous sampling time The winding radius is the starting value for the recursion; The structural proportionality coefficient describes the relationship between how much the yarn length increases and how much the radius increases. It is related to the yarn bobbin structure and is a constant obtained through experimental data calibration. The previous sampling time The linear velocity.

[0111] S603: Based on the instantaneous angular velocity and winding radius, the instantaneous linear velocity at each sampling moment is calculated and merged to construct an original linear velocity sequence, which reflects the actual speed of the yarn as it passes through the measuring aperture. In the formula, Let n be the linear velocity at the nth sampling time. Angular velocity at the nth sampling time.

[0112] S604: By performing a moving average smoothing on the original linear velocity sequence, a less noisy and more gradually changing effective linear velocity sequence is obtained, which is used to provide a stable and reliable velocity input for subsequent time-length mapping, i.e.: In the formula, M is the effective linear velocity after smoothing; M is the sliding window length. This represents the original linear velocity value at the m-th sampling point counting backwards from the current moment.

[0113] S605: By integrating and accumulating the effective linear velocity sequence over time, the cumulative through length corresponding to each time sampling point is calculated, which is used to establish the mapping relationship from the time axis to the physical length axis; that is: In the formula, The initial length position is set to 0, representing the length reference point at the start of the measurement; The cumulative length of the yarn relative to the initial point at the nth sampling time; Let N be the linear velocity at the i-th sampling moment in the effective linear velocity sequence; N is the total number of samples in the effective linear velocity sequence.

[0114] S606: Based on dividing the cumulative length into equidistant length positions at fixed intervals, a set of regular cumulative length sequences is obtained, which are used as the target length coordinate system after resampling; that is:

[0115]

[0116] In the formula, This indicates the length position of the kth target length sampling point, starting from 0 and increasing by a fixed length step each time. The length sampling interval determines the resolution on the length axis after resampling; k is the index number of the length position; K is the total number of equally spaced length sampling points, i.e. ;in The total length of the yarn that has passed through up to the last time sampling point is the endpoint of the cumulative effective linear velocity sequence; This indicates rounding down to the nearest integer.

[0117] S607: For each target length sampling point in the cumulative length sequence Find an index in the cumulative length sequence , making The length corresponding to two adjacent time sampling points and Between, that is, to determine which two time measurement points this length position falls exactly between: Then, by performing linear interpolation between adjacent time samples according to length position, the final line density sequence at corresponding equidistant length positions is obtained. This sequence is used to form a line density signal that is equally spaced on the length axis and can be directly used for statistical and frequency domain analysis, i.e.:

[0118]

[0119] In the formula, For the length position Linear density after tension compensation; For at a certain point in time The starting point of the interpolation interval is the linear density after tension compensation. For at a certain point in time The measured and tension-compensated linear density is the endpoint of the interpolation interval; ultimately, a linear density sequence with equal length intervals is obtained. Therefore, by mapping the true linear density sequence sampled at equal intervals to the actual physical length coordinate system through the effective linear velocity sequence, and constructing a linear density sequence with equal length intervals using linear interpolation, the problem of non-uniform sampling point density caused by velocity fluctuations can be eliminated. That is, the true linear density sequence can be equivalent to the final linear density sequence, i.e.: .

[0120] S7: Based on the final linear density sequence, which has been unified to length coordinates and under reference tension and position conditions, calculate the average linear density value over the entire measured length range, and use this as the overall linear density level of the yarn for that segment, i.e.: In the formula, This represents the overall average linear density of the yarn over a specific measurement length range; This represents the linear density at the k-th point of equal length interval on the yarn.

[0121] S8: Further, based on the construction of a zero-mean fluctuation sequence, the DC component is avoided from interfering with the energy analysis of the non-uniform fluctuation in the spectrum, i.e.: In the formula, This represents the amount of fluctuation in the linear density of the yarn at the k-th point in a zero-mean fluctuation sequence, which deviates from the overall average value.

[0122] However, in actual production, due to the imprecise temperature control of the hot rollers in spinning machines, the surface temperature of the hot rollers fluctuates periodically. This temperature fluctuation is transmitted to the yarn, causing the linear density of the yarn to drift linearly along its length during stretching, resulting in a slow and continuous change in the yarn's physical state. Consequently, the calculated zero-mean fluctuation sequence will contain this linear trend component, leading to an overestimation of the standard deviation and yarn unevenness in subsequent calculations. This results in a misjudgment of yarn quality, attributing systematic errors caused by the process to the inherent unevenness of the yarn, thus reducing the accuracy and reliability of the assessment.

[0123] Therefore, by detecting and removing linear drift in the zero-mean fluctuation sequence, a pure sequence containing only random fluctuation components is generated, ensuring that the yarn unevenness truly reflects the inherent quality fluctuations of the yarn, unaffected by external process interference. Specifically:

[0124] S801: By summing the cumulative length sequences, the sum of length positions is obtained, which is used as the basic statistic in subsequent linear regression calculations, i.e.: In the formula, It represents the sum of all length positions, describing the overall distribution of length coordinates.

[0125] S802: By summing the squares of the cumulative length sequences, the sum of the squares of the length positions is obtained, which is used to calculate the denominator of the linear trend slope, i.e.: In the formula, This represents the sum of the squared values ​​of all length positions, reflecting the dispersion of the length coordinates. It is used to ensure the stability of slope calculation and prevent calculation overflow caused by an excessively small denominator.

[0126] S803: By summing the final linear density sequences, the sum of linear densities is obtained and used as the statistic of the dependent variable in linear regression; that is: In the formula It represents the sum of all linear density values ​​within the entire measurement length.

[0127] S804: By summing the product sequences of length position and linear density, a sum parameter of their products is obtained, which is used to calculate the numerator of the linear trend slope; that is: In the formula It represents the sum of the covariance between length and linear density, quantifying the cooperative variation relationship between length and linear density.

[0128] S805: Calculate the slope of the yarn's linear trend based on the sum of the length positions, the sum of the squared length positions, the sum of the linear density values, and the sum of the linear density covariance relationships, i.e.: In the formula, A is the slope of the linear trend of the yarn, which represents the rate at which the linear density of the yarn changes with unit length.

[0129] S806: Calculate the intercept of the yarn linear trend based on the slope of the yarn linear trend, the sum of the linear density values, and the sum of all length positions, i.e.: In the formula, B is the intercept of the linear trend of the yarn, which represents the linear density value of the fitted linear trend at the starting point on the length coordinate.

[0130] S807: For each sample point in the final linear density sequence, calculate the detrended linear density value to generate a detrended linear density sequence to reflect the inherent fluctuations of the yarn; that is: In the formula, Indicates the position at length The detrended line density at a given point is the line density value after removing the effects of linear drift.

[0131] S808: Update the zero-mean fluctuation sequence based on the detrended line density sequence. Since the mean of the detrended line density sequence is already zero, the detrended line density sequence is directly taken as the detrended value. .

[0132] S9: Calculate the overall linear density variance and standard deviation under unified physical conditions based on zero-mean fluctuation sequences, i.e.: In the formula, This represents the overall linear density variance under uniform physical conditions; It represents the standard deviation of the overall linear density under uniform physical conditions.

[0133] S10: Calculate the overall strip unevenness rate based on the overall linear density variance and standard deviation under uniform location, uniform tension, and uniform length coordinates, i.e.: In the formula, The corrected overall yarn unevenness indicates that the unevenness was calculated under a uniform length coordinate and a uniform reference tension. This means that all measurement errors introduced by dynamic changes in the equipment (such as speed fluctuations, tension changes, and yarn sway) have been eliminated, and the result purely reflects the quality fluctuation of the yarn itself.

[0134] An intelligent quality assessment system for spun yarn products includes a capacitance signal acquisition module, a sum and difference signal construction module, a linear relationship establishment module, an inverse operation module, a position correction module, a mapping module, a calculation module, and a sequence generation module. Wherein:

[0135] The capacitance signal acquisition module is responsible for acquiring raw capacitance signals in real time from multiple electrodes (channels) of the capacitance evenness meter. First, it reads the reference capacitance value when no yarn passes through to eliminate zero-point drift inherent in the environment and circuitry. Next, it standardizes the acquired raw capacitance signal using a calibrated gain factor; that is, it subtracts the reference value from the raw signal and divides by the gain factor, ultimately outputting a standardized capacitance signal that is directly and stably related to the physical quantity. This provides a high-quality, comparable input signal for all subsequent processing. By eliminating interference from the equipment itself and the environment, it ensures that data measured at different times and through different channels are based on a unified reference, laying a solid foundation for accurate physical quantity inversion.

[0136] The sum and difference signal construction module receives multi-channel (taking a dual-channel example) standardized capacitance signals from the capacitance signal acquisition module. It uses specific linear combination operations to add the signals from the two channels to obtain the sum signal and subtract them to obtain the difference signal. This initially separates the contributions of different physical effects at the signal level. The sum signal is sensitive to changes in the overall linear density (mass) of the yarn, but relatively insensitive to small lateral oscillations of the yarn in an electric field; while the difference signal is highly sensitive to changes in the lateral position of the yarn. This construction creates the conditions for subsequent separate calculations of linear density and position information.

[0137] The linear relationship establishment module applies a pre-determined relationship matrix established through calibration experiments. This linear model describes the mathematical relationship between the sum and difference signals and the actual physical quantities of the yarn (linear density and lateral position), establishing a reliable bridge connecting electrical signals and the physical world. Through this calibrated linear model, the system can quantitatively interpret changes in the measured capacitance signal as changes in the physical state of the yarn, a crucial step in achieving signal-to-information conversion.

[0138] The inverse operation module receives the sum and difference signals from the sum and difference signal construction module in real time. It utilizes the model provided by the linear relationship construction module to perform inverse matrix operations. By solving a set of linear equations, it decouples and calculates two independent sequences of physical quantities from the sum and difference signals: one is the instantaneous linear density representing yarn thickness, and the other is the instantaneous lateral position representing the yarn's left-right offset within the measuring groove. This decoupling of physical quantities separates the superimposed linear density and position information, thus obtaining preliminary, independent linear density estimates and accurate position information, providing direct input for subsequent error compensation.

[0139] The position correction module receives the instantaneous linear density sequence and lateral position sequence output by the inverse operation module. It calls a pre-calibrated position correction function, which describes the linear density measurement error caused by the yarn deviating from the electric field center. Using the current lateral position, the module calculates a correction factor through this function to compensate for the initially estimated linear density value in real time. This eliminates the nonlinear error in the electric field caused by the periodic oscillation of the yarn within the measuring slot. Through this correction, the linear density value measured at any position is equivalently converted to the linear density value when the yarn is at its ideal center position, significantly improving the accuracy of linear density measurement.

[0140] The mapping module estimates and smooths the instantaneous linear velocity sequence of the yarn passing through the measurement point using an angular velocity sensor and a winding radius model. Then, by integrating the linear velocity, the yarn length coordinates corresponding to each sampling time point are obtained. Finally, through resampling technology, the linear density sequence, which is equally spaced on the time axis but non-uniform in physical length, is converted into a strictly equally spaced linear density sequence on the length axis. This solves the problem that equally spaced time sampling is not equal to equally spaced length sampling due to fluctuations in winding linear velocity. By unifying to physical length coordinates, it ensures that all subsequent quality indicators are calculated based on a true length benchmark, making the results conform to industry standard definitions and comparable under different working conditions.

[0141] The sequence generation module receives the length-domain equally spaced linear density sequence output by the mapping module. First, it detects and removes any potential linear trends from this sequence, generating a clean, detrended sequence that contains only the random fluctuations of the yarn itself. Then, it subtracts the mean from this sequence to generate a fluctuation sequence with a mean of zero. This process eliminates systematic drift caused by factors other than the yarn's structural characteristics before calculating the overall fluctuation. This ensures that the subsequently calculated fluctuation index accurately and purely reflects the yarn's inherent quality non-uniformity, avoiding misjudgments in quality assessment results due to process interference.

[0142] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the protection scope of the technical solutions of the embodiments of the present invention.

Claims

1. A smart quality assessment method for spun yarn products, characterized in that, Includes the following steps: S1: Acquire multi-channel capacitance signals and perform preprocessing to obtain standardized capacitance signals; S2: Construct a sum-difference capacitance signal based on the standardized capacitance signal; S3: Establish a linear relationship between the sum and difference capacitance signals and the yarn linear density and lateral position; S4: Based on the aforementioned linear relationship, the instantaneous linear density sequence and the lateral position sequence are obtained through inverse operation; S5: Perform position correction on the instantaneous linear density sequence to obtain the compensated linear density sequence; S6: Map the compensated line density sequence from the time domain to the length domain to obtain a line density sequence with equal length intervals; S7: Calculate the average line density based on the equally spaced line density sequence; S8: Generate a zero-mean fluctuation sequence based on the average linear density; S9: Calculate the linear density variance and standard deviation based on the zero-mean fluctuation sequence; S10: Calculate the strip unevenness rate based on the standard deviation and average linear density.

2. The intelligent quality assessment method for spun yarn products according to claim 1, characterized in that, The preprocessing described in S1 includes: Calculate the normalized capacitance signal In the formula, The standardized capacitance signal of the i-th channel at time t; Let be the original capacitance value of the i-th channel measured at time t; This is the reference capacitance of the i-th channel when no yarn passes through; is the calibration gain factor for the i-th channel; i is the index number.

3. The intelligent quality assessment method for spun yarn products according to claim 2, characterized in that, The sum and difference capacitance signals mentioned in S2 include sum signals. Sum and difference signals ,in: In the formula and These are the standardized capacitor signals for the first and second channels, respectively.

4. The intelligent quality assessment method for spun yarn products according to claim 3, characterized in that, The linear relationship described in S3 is expressed as follows: In the formula, Let be the linear density of the yarn passing through the electrode cross-section at time t; Let t be the transverse position of the yarn within the electrode cross-section; For constant terms; The sensitivity coefficient of the signal to line density; Let be the residual sensitivity coefficient of the signal to lateral position; The sensitivity coefficient of the difference signal to the line density; The residual sensitivity coefficient of the difference signal to the lateral position.

5. The intelligent quality assessment method for spun yarn products according to claim 4, characterized in that, The inverse operation described in S4 includes: The instantaneous linear density sequence and the lateral position sequence are obtained by inverse operation of the linear mapping matrix, wherein... M is the calibrated linear mapping matrix.

6. The intelligent quality assessment method for spun yarn products according to claim 5, characterized in that, The position correction described in S5 includes: The instantaneous linear density sequence is corrected using a position correction function to obtain the compensated linear density sequence. ,in ; For the reference gain term, For linear position offset terms, These are quadratic nonlinear terms; all were obtained through calibration using standard yarns.

7. The intelligent quality assessment method for spun yarn products according to claim 6, characterized in that, S5 further includes tension compensation for the position correction of the instantaneous linear density sequence, the tension compensation including: S501: Acquires the tension output signal through a tension sensor and calibrates to obtain the true tension sequence, i.e.: In the formula, For true tension; This is the slope calibration coefficient for the linear relationship between tension and sensor output; This is the intercept calibration coefficient in the linear relationship between tension and sensor output; S502: Discretely sample the real tension sequence to obtain the instantaneous tension sequence, i.e.: ;in, The instantaneous tension value at the nth sampling time; For a moment The corresponding actual tension; For at any time The tension output signal; S503: Smooth the instantaneous tension sequence to obtain a tension estimate sequence, i.e.: In the formula, Let be the tension estimate at the nth sampling time; Q is the window length of the sliding process; q is the index number of the window. Let be the instantaneous tension value at the nq-th sampling time; S504: Calculate the yarn elongation ratio based on the tension estimation value sequence, i.e.: In the formula, This represents the actual length of the yarn segment under the current tension estimate. This is the actual length of the yarn segment under the reference tension; For the strain, i.e.: ;in For additional elongation; As the reference tension; The equivalent compliance within a given tension range; S505: Calculate the apparent linear density based on the elongation ratio, i.e.: In the formula, The apparent linear density of the yarn of the current length under tension T; The baseline linear density under reference tension; S506: Based on the apparent linear density and reference tension, calculate the true linear density sequence, i.e.: In the formula, This is the actual linear density under the reference tension, derived from the apparent linear density under the current tension.

8. The intelligent quality assessment method for spun yarn products according to claim 7, characterized in that, S6 further includes mapping the compensated linear density sequence from the time domain to the length domain: S601: Collect the instantaneous angular velocity sequence of the winding roller; S602: Based on the instantaneous angular velocity sequence and the winding radius recursive model, update the winding radius sequence, i.e.: In the formula, The winding radius at the nth sampling time; The previous sampling time The winding radius; This is the structural proportionality coefficient; The previous sampling time linear velocity; S603: Based on the winding radius sequence and instantaneous angular velocity sequence, calculate the original linear velocity sequence, i.e.: In the formula, Let n be the linear velocity at the nth sampling time. The angular velocity at the nth sampling time; S604: Smooth the original linear velocity sequence to obtain an effective linear velocity sequence, i.e.: In the formula, M is the effective linear velocity after smoothing; M is the sliding window length. This represents the original linear velocity value at the m-th sampling point counting backwards from the current moment; S605: Calculate the cumulative passing length sequence based on the effective linear velocity sequence; that is: In the formula, This is the initial length position; The cumulative length of the yarn relative to the initial point at the nth sampling time; Let N be the linear velocity at the i-th sampling time in the effective linear velocity sequence; N is the total number of samples in the effective linear velocity sequence. S606: Based on the aforementioned cumulative length sequence, divide the equidistant length position sequence, that is: In the formula, This indicates the length position of the k-th target length sampling point; is the length sampling interval; k is the index number of the length position; K is the total number of equally spaced length sampling points, i.e. ;in This refers to the total length of the yarn that has passed up to the last time sampling point; Indicates rounding down; S607: The compensated linear density sequence is mapped to the equidistant length position sequence through linear interpolation to obtain a linear density sequence with equal length intervals, i.e.: In the formula, For the length position Linear density after tension compensation; For at a certain point in time The linear density after tension compensation was measured. For at a certain point in time The linear density is measured and compensated for by tension; finally, a linear density sequence with equal length intervals is obtained.

9. The intelligent quality assessment method for spun yarn products according to claim 8, characterized in that, S8, generating a zero-mean fluctuation sequence based on the average linear density, further includes removing linear drift, which includes: S801: Calculate the sum of the length position sequence ; S802: Calculate the sum of the length position square sequence. ; S803: Calculate the sum of the equally spaced linear density sequences. ; S804: Calculate the sum of the product sequence of length position and linear density. ; S805: Based on , , and Calculate the slope of the linear trend ; S806: Based on , Calculate the linear trend intercept with A. ; S807: Based on A and B, detrend the line density sequences with equal length intervals to obtain a detrended line density sequence, i.e.: In the formula, Indicates the position at length Detrending line density at the location; S808: Use the detrended linear density sequence as the updated zero-mean fluctuation sequence.

10. A system for implementing the intelligent quality assessment method for spun yarn products according to any one of claims 1-9, characterized in that, include: The capacitor signal acquisition module is used to acquire multi-channel capacitor signals and perform preprocessing. The sum and difference signal construction module is used to construct sum and difference capacitance signals; The linear relationship establishment module is used to establish the linear relationship between the capacitance signal and the physical quantity; The inverse operation module is used to calculate the instantaneous linear density sequence and the lateral position sequence; The position correction module is used to correct the position of the line density; The mapping module is used to map the time domain to the length domain; The sequence generation module is used to generate zero-mean fluctuation sequences.

Images