Tension detection method and system for yarn conveying

Abstract

The invention belongs to the technical field of data processing methods, and particularly relates to a tension detection method and system for yarn conveying to solve the technical problem that a detection method in the prior art is low in precision, and the method comprises the following steps: S1, collecting multi-modal vibration data of a yarn in a preset detection area, the multi-modal vibration data comprises a yarn vibration displacement time sequence image obtained based on a linear array camera and a vibration acoustic signal obtained based on an acoustic sensor; s2, constructing a space-time joint feature vector representing the vibration dynamic characteristics of the yarn; s3, separating out pure yarn vibration characteristics influenced by background vibration of the rejecting equipment; and S4, determining the instantaneous tension of the yarn by solving the inverse problem of the nonlinear wave equation coupled with the tension, the bending moment and the air damping. The linear array camera and the acoustic sensor are adopted, multi-mode and multi-dimensional information collection of yarn vibration is achieved, space-time joint feature vectors are constructed, and complex vibration behaviors of yarn can be represented more comprehensively and accurately.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of data processing method, and particularly relates to a yarn conveying tension detection method and system. BACKGROUND

[0002] In the yarn processing, winding and weaving links of the textile industry, yarn tension is a crucial process parameter. Accurate and real-time monitoring and control of yarn tension is the basis for ensuring the quality of the final product, reducing yarn breakage, improving production efficiency and achieving automated production. Uneven tension can cause defects such as weft skew, streaks and uneven fabric surface, and excessive tension can easily cause yarn breakage, affecting the continuity of production. Therefore, it is of great practical significance to develop a high-precision and high-reliability yarn tension online detection method.

[0003] Currently, yarn tension detection methods are mainly divided into contact and non-contact types. Traditional contact methods, such as three-roller tension sensors, measure the pressure on the guide rollers to indirectly reflect the tension by winding the yarn around a set of guide rollers. Although this method is mature, it has obvious drawbacks: the sensor directly contacts the high-speed moving yarn, which generates additional friction, not only interfering with the tension measurement value itself, but also potentially damaging the yarn surface, especially for high-count high-density or fragile special fibers. In addition, the wear and inertia of mechanical components also limit their response speed and long-term stability. In order to overcome these shortcomings, non-contact detection methods based on vibration principles have emerged. This method simplifies the yarn as a vibrating string, and according to the string vibration theory, its natural frequency is proportional to the square root of the tension value. By measuring the vibration frequency of the yarn, its tension can be inferred. However, existing non-contact vibration methods still face great challenges in practical application. On the one hand, the textile workshop environment is complex, and equipment such as motors, spindles and guide rails can produce strong background vibrations and noise. These interference signals will be coupled with the weak yarn vibration signals, resulting in a very low signal-to-noise ratio, making it difficult to accurately extract the pure yarn eigenfrequency, thereby seriously affecting the measurement accuracy. On the other hand, traditional vibration models often idealize the yarn as a flexible string with no stiffness and linear elasticity, ignoring the actual physical characteristics such as the bending stiffness of the yarn itself, air damping effect and material nonlinearity. This over-simplification leads to a large deviation between the theoretical model and the actual working condition, especially at high speed and high tension, the error is more significant. In addition, the single modal sensing method (such as using only a laser displacement sensor or a microphone) has limited information dimension, making it difficult to fully characterize the complex spatiotemporal vibration behavior of the yarn, limiting the further improvement of detection accuracy. SUMMARY

[0004] The present application provides a yarn conveying tension detection method and system to solve the technical problem of low precision of the detection method of the prior art.

[0005] To solve the above problems, the yarn tension detection method for yarn conveying provided by the present application adopts the following technical scheme: a yarn tension detection method for yarn conveying, comprising the following steps: S1, collecting multi-modal vibration data of the yarn in a preset detection area, the multi-modal vibration data comprising yarn vibration displacement time sequence images obtained based on a line array camera and vibration acoustic signals obtained based on an acoustic sensor; S2, based on the yarn vibration displacement time sequence images, calculating the velocity field of at least two spatially separated measuring points on the yarn using an optical flow method, and combining the frequency spectrum analysis result of the vibration acoustic signals to construct a space-time joint feature vector representing the vibration dynamics characteristics of the yarn; S3, using a device systematic interference model associated with the yarn conveying speed to decouple and calculate the space-time joint feature vector to separate out pure yarn vibration characteristics excluding the influence of device background vibration; S4, based on the pure yarn vibration characteristics and in combination with preset constitutive parameters of the yarn, solving the inverse problem of a nonlinear wave equation coupled with tension, bending moment and air damping to determine the instantaneous tension of the yarn, the preset constitutive parameters including the linear density, bending stiffness and nonlinear elastic coefficient of the yarn.

[0006] Further, in S1, the line array camera is used to capture the main vibration mode of the yarn at a capture frame rate; the acoustic sensor is arranged in the preset detection area so that the acoustic sensor can pick up the vibration acoustic signals generated by the yarn vibration Further, in S1, the line array camera is arranged vertically opposite the yarn.

[0007] Further, in S2, a plurality of measuring points distributed along the yarn conveying path are selected in the vibration displacement time sequence images, and an optical flow algorithm is used to calculate the vibration velocity time sequence of each measuring point perpendicular to the yarn conveying direction.

[0008] Further, in S2, the vibration acoustic signals are subjected to frequency spectrum analysis to extract one or more main frequency characteristics representing the main energy distribution thereof; the main frequency characteristics are fused with the vibration velocity time sequence characteristics of the plurality of measuring points to form the space-time joint feature vector.

[0009] Further, in S3, during decoupling calculation, the device systematic interference model is constructed as a state space model, parameters of the state space model being determined by calibrating the vibration characteristics of the device when running empty at a specific conveying speed; a filter-based state estimation algorithm is used to take the space-time joint feature vector as an observation input to estimate and separate out the pure yarn vibration characteristics therefrom.

[0010] Further, in S4, the non-linear wave equation takes into account the bending stiffness of the yarn itself and the air damping effect received during movement.

[0011] Further, in S4, a target function is constructed with the instantaneous tension value of the yarn as the parameter to be optimized, and the error minimization between the pure yarn vibration characteristics and the vibration characteristics predicted by the non-linear wave equation as the target; an iterative optimization algorithm is used to solve the optimization problem to obtain the instantaneous tension value.

[0012] Further, the error between the pure yarn vibration characteristics and the vibration characteristics predicted by the non-linear wave equation is the root mean square error.

[0013] The application also provides a yarn conveying tension detection system, comprising a memory and a processor, the memory stores computer program instructions, when the computer program instructions are executed by the processor, the yarn conveying tension detection method of any one of the above is realized.

[0014] The beneficial effect is: compared with the prior art, the application realizes multi-modal and multi-dimensional information acquisition of yarn vibration by jointly using a linear array camera and an acoustic sensor, and constructs a space-time joint feature vector, which can more comprehensively and accurately represent the complex vibration behavior of the yarn, and overcomes the defect of insufficient information dimension of a single sensing mode. At the same time, by introducing a device systematic interference model and decoupling calculation, the interference of strong background vibration noise can be effectively stripped, and the pure yarn vibration characteristics are extracted, which significantly improves the anti-interference ability and reliability in complex industrial environments. A non-linear wave equation coupled with the bending stiffness of the yarn, air damping and non-linear elastic coefficient is used to solve the inverse problem, which is closer to the real physical state of the yarn at high speed, and fundamentally ensures the physical authenticity and accuracy of the tension calculation result, solving the technical problem of large measurement error caused by model simplification in the prior art. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 The flowchart of the yarn conveying tension detection method; Figure 2 The structural block diagram of the yarn conveying tension detection system. DETAILED DESCRIPTION

[0016] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application, and those skilled in the art should know that the embodiments described below are part of the embodiments of the present disclosure, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the present application.

[0017] Embodiments of the yarn conveying tension detection method provided by the present invention: like Figure 1 As shown, the yarn feeding tension detection method includes the following steps: S1, collecting multimodal vibration data of the yarn in a preset detection area, the multimodal vibration data including the yarn vibration displacement time-series image obtained by the linear array camera and the vibration acoustic signal obtained by the acoustic sensor.

[0018] Specifically, the linear array camera is deployed perpendicular to the yarn movement path, and the vibration profile of the yarn in a fixed length segment is continuously captured at a frame rate of more than one thousand frames per second, forming a two-dimensional yarn vibration displacement time-series image in which the grayscale value changes over time; at the same time, a high-sensitivity micro-electromechanical system microphone is arranged non-contact near the detection area to synchronously collect the vibration acoustic signal radiated to the surrounding air when the yarn vibrates at a sampling rate of 48kHz.

[0019] In an optional embodiment, a linear array camera is used to shoot at an acquisition frame rate capable of capturing the main vibration modes of the yarn; and an acoustic sensor is arranged in a preset detection area so that the acoustic sensor can pick up the vibration acoustic signal generated by the yarn vibration.

[0020] Specifically, to clearly capture the vibrational patterns of yarns in high-speed motion, a high-frequency linear array camera (such as the racer series from Basler, Germany) is selected, with an acquisition frame rate set to no less than 2000 Hz. Considering that the primary vibration frequencies of yarns typically range from a few hundred Hz to 1000 Hz, and based on the Nyquist theorem, a 2000 Hz acquisition frame rate is sufficient to avoid signal aliasing and ensure distortion-free reconstruction of the vibration waveform at each point on the yarn, thus providing a high-quality raw image sequence for subsequent velocity field calculations.

[0021] The acoustic sensor is a highly sensitive micro-electromechanical system microphone that captures vibroacoustic signals coupled to yarn vibrations, complementing the visual information. The acoustic sensor is mounted approximately 5 cm from the yarn's average path, with its pickup direction oriented directly toward the center of the area with the highest yarn vibration amplitude. This distance ensures sufficient sound pressure signals generated by yarn vibrations while effectively minimizing airflow noise interference caused by the yarn's high-speed motion. This maximizes the signal-to-noise ratio and enables the detection of vibroacoustic signals closely related to tension fluctuations.

[0022] S2, based on the time-series image of yarn vibration displacement, uses the optical flow method to calculate the velocity field of at least two spatially separated measuring points on the yarn, and combines the spectrum analysis results of the vibroacoustic signal to construct a spatiotemporal joint feature vector that characterizes the vibration dynamics of the yarn.

[0023] Two fixed column pixels representing yarn position are selected as measuring points in the yarn vibration displacement time series images. Lucas-Kanade optical flow estimation algorithm is applied to calculate the pixel motion velocity of each measuring point in the direction perpendicular to the yarn axial direction, and two groups of pixel motion velocity time series are obtained. Short-time Fourier transform is performed on the synchronously collected vibration acoustic signal to extract the main harmonic frequency and its energy amplitude in the time-frequency spectrum. Finally, the instantaneous velocity values of the two measuring points are concatenated with the main harmonic frequency and energy amplitude of the vibration acoustic signal to form a spatio-temporal joint feature vector containing velocity information and frequency domain information.

[0024] In an alternative embodiment, when calculating the velocity field of at least two spatially separated measuring points on the yarn, multiple measuring points distributed along the yarn conveying path are selected in the yarn vibration displacement time series images, and an optical flow-based algorithm is used to calculate the vibration velocity time series of each measuring point in the direction perpendicular to the yarn conveying direction.

[0025] Specifically, in the sequence of continuous image frames obtained by the line array camera, the yarn image is equally divided into four regions along its length direction, and the center positions of each region, i.e., the 256th, 512th, 768th, and 1024th pixel columns along the yarn path, are set as fixed measuring points. Multiple spatially separated measuring points, such as these four points, are selected to observe the propagation characteristics of the vibration wave on the yarn, rather than just the vibration state of a single point, which provides rich spatial information for subsequent more accurate inverse analysis of physical model parameters.

[0026] For each selected measuring point, the Horn-Schunk algorithm in the dense optical flow algorithm is used to calculate its velocity perpendicular to the yarn conveying direction. By analyzing the gray level changes of the pixels in the neighborhood of each measuring point between two consecutive image frames, such as the tth frame and the t+1th frame, the motion vector of the pixel is estimated. Since the yarn mainly vibrates in the direction perpendicular to the yarn conveying direction, the vertical component of the motion vector is extracted and converted to physical velocity in m / s according to the pixel size and frame rate of the line array camera. Repeating this process for all image frames, the complete data sequence of the vibration velocity of each measuring point as a function of time is obtained.

[0027] In an alternative embodiment, when constructing the spatio-temporal joint feature vector, the vibration acoustic signal is subjected to spectral analysis to extract one or more main frequency features representing its main energy distribution; the main frequency features are fused with the vibration velocity time series features of multiple measuring points to form the spatio-temporal joint feature vector.

[0028] Specifically, the raw vibroacoustic signal collected by the acoustic sensor is preprocessed and then subjected to time-frequency analysis using the short-time Fourier transform method. The signal is segmented into 100ms window lengths and 50ms step sizes, and the power spectral density (PSD) of each segment is calculated. Within the resulting spectrum, the two frequency peaks with the highest energy are identified. For example, at a specific moment, the dominant frequency is identified as 450Hz, and the secondary frequency is identified as 910Hz. These two frequency values, as the acoustic signature within that time window, reflect the yarn's most significant resonance mode at that moment.

[0029] Within the same time window, statistical features are extracted from the vibration velocity time series at the four measurement points. For example, the root mean square value and kurtosis coefficient of the vibration velocity signal at each measurement point are calculated, resulting in a total of eight visual features. The two acoustic features extracted in the previous step are then concatenated with these eight visual features to form a ten-dimensional vector. This vector is the spatiotemporal joint feature vector at that moment. It combines the frequency domain information of the vibroacoustic signal with the temporal and spatial distribution information of the visual signal. Compared to a single modal feature, it can more comprehensively characterize the complex vibration state of the yarn.

[0030] S3, using the equipment systematic interference model associated with the yarn delivery speed, decouples the spatiotemporal joint feature vector to separate the pure yarn vibration characteristics that eliminate the influence of equipment background vibration.

[0031] The equipment's systematic interference model is a blind source separation matrix built using an independent component analysis algorithm. When the equipment is unloaded or running a baseline non-vibrating yarn, background vibration data at different conveying speeds is collected and a spatiotemporal joint eigenvector is constructed. A fast independent component analysis algorithm is then used to train an unmixing matrix capable of separating background vibration sources. During actual testing, the real-time spatiotemporal joint eigenvector is multiplied by the unmixing matrix to isolate the independent components representing pure yarn vibration from the mixed signal, reconstructing the pure yarn vibration signature.

[0032] During the decoupling calculation, the equipment systematic interference model is constructed as a state-space model. The parameters of this model are determined by calibrating the vibration characteristics of the equipment when it is running no-load at a specific conveying speed. A filtering-based state estimation algorithm is adopted, and the joint time-space eigenvector is used as the observation input to estimate and separate the pure yarn vibration characteristics.

[0033] Specifically, the systemic interference of the equipment mainly comes from rotating parts such as motors, bearings and guide wheels, whose vibration has a strong periodicity. The standard production speed of the device is run, while the vibration signals of the device frame and key components are collected by using acceleration sensors and line array cameras. By performing spectral analysis on these idle signals, the main interference frequencies are identified, such as the 50Hz motor power frequency interference and the 180Hz guide wheel rotation frequency interference, and a second-order state space model describing the linear superposition of these interference sources is constructed based on this, and the model parameters are determined.

[0034] The extended Kalman filtering algorithm is used to realize the decoupling of the interference signal and the yarn signal. The device systematic interference model constructed in the previous step is used as a prediction model to predict the interference state at the next moment. The ten-dimensional spatio-temporal joint feature vector constructed in the previous step is used as an observation vector input into the extended Kalman filter. The extended Kalman filter compares and corrects the observed spatio-temporal joint feature vector with the model predicted interference through continuous iteration of the prediction and update steps, thereby optimally estimating the part contributed by the device interference and subtracting it from the original observation, and finally outputting a feature vector containing only the pure yarn vibration information.

[0035] S4, based on the pure yarn vibration feature and combined with the preset constitutive parameters of the yarn, the inverse problem of the nonlinear wave equation coupled with tension, bending moment and air damping is solved to determine the instantaneous tension of the yarn, and the preset constitutive parameters include the linear density, bending stiffness and nonlinear elastic coefficient of the yarn.

[0036] A fourth-order partial differential equation based on the Euler-Bernoulli beam theory and introducing a nonlinear tension term and a linear air damping term is established as a yarn vibration model; the finite difference method is used to discretize the fourth-order partial differential equation to construct a forward model, which can predict the vibration frequency and amplitude of the yarn according to the given tension value; finally, the measured pure yarn vibration feature is taken as the target, and the Levenberg-Marquardt optimization algorithm is used to iteratively adjust the tension value input to the forward model, so that the sum of squared residuals between the predicted vibration feature and the measured feature is minimized, and when the residuals converge, the corresponding tension value is the finally determined instantaneous tension of the yarn.

[0037] In an optional embodiment, the nonlinear wave equation takes into account the bending stiffness of the yarn itself and the air damping effect during movement.

[0038] Specifically, the traditional ideal flexible string model ignores the bending stiffness of the yarn, which will introduce a large error when dealing with high twist or high denier yarns. Therefore, a term proportional to the fourth-order partial derivative of displacement with respect to spatial coordinates is added to the nonlinear wave equation, i.e. , where is the Young's modulus of the yarn, is the cross-sectional moment of inertia of the yarn, is the fourth-order partial derivative of displacement with respect to spatial coordinates. For example, for a 150-denier polyester industrial yarn, its bending stiffness is The value of is approximately The introduction of this term enables the model to more accurately describe the mechanical behavior of yarn under high-frequency vibration or small curvature radius.

[0039] When the yarn moves and vibrates at high speed in the air, it will be subject to the viscous resistance and disturbance resistance of the air, which will dissipate the vibration energy and cause vibration attenuation. In order to quantify this effect, a damping term proportional to the first-order partial derivative of the yarn vibration velocity, i.e., displacement with respect to time, is added to the nonlinear wave equation, i.e. ,in is the air damping coefficient, is the first-order partial derivative of displacement with respect to time. Through wind tunnel experiments or computational fluid dynamics simulations, the At the conveying speed, the damping coefficient The value of is approximately After accounting for air damping, the model's predictions can better match the amplitude attenuation characteristics of the measured vibration signal.

[0040] In an optional embodiment, when solving the inverse problem, an objective function is constructed with the instantaneous tension of the yarn as the parameter to be optimized and the goal of minimizing the error between the vibration characteristics of the pure yarn and the vibration characteristics predicted by the nonlinear wave equation; and an iterative optimization algorithm is used to solve the optimization problem to obtain the instantaneous tension value.

[0041] Specifically, the solution to the inverse problem is formulated as an optimization task. The objective function is defined as the root mean square error (RMS) between the measured pure vibration characteristics and the vibration characteristics predicted by the physical model. Specifically, the RMS error between the pure vibration velocity time series of the four measuring points obtained through decoupling and the predicted velocity time series of the four corresponding measuring points obtained by solving the nonlinear wave equation using numerical methods such as the finite difference method under the assumption of a given instantaneous tension T is calculated. This RMS error is a function of the parameter to be optimized, T.

[0042] To solve this optimization problem, a Gauss-Newton iterative algorithm is used. An initial guess for the tension is required, such as 15 cN, which is empirically set. In each iteration, the Jacobian matrix of the objective function is calculated based on the current tension value, and then a linear least-squares problem is solved to obtain the updated tension value. For example, after the first iteration, the tension value may be updated to 15.3 cN. This iterative process continues until the difference between the tension values ​​calculated in two consecutive iterations is less than a preset convergence threshold, such as 0.01 cN, or the decrease in the objective function is less than a minimum value. At this point, the tension value is considered to be the optimal estimate of the instantaneous tension of the yarn at the current moment.

[0043] Embodiments of the yarn conveying tension detection system provided by the present application: As shown in Figure 2 The yarn conveying tension detection system includes a processor and a memory storing computer program instructions that, when executed by the processor, implement the yarn conveying tension detection method described above.

[0044] The yarn conveying tension detection system also includes a communication interface and other components well known to those skilled in the art, the settings and functions of which are known in the art, and therefore will not be described here.

[0045] In the present application, the aforementioned memory can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, the computer readable storage medium can be any suitable magnetic storage medium or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random-Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store the desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of the device or accessible or connectable to the device. Any application or module described by the present application can be implemented using computer readable / executable instructions stored or otherwise held by such computer readable medium.

[0046] In addition, in the description of the present specification, the meaning of "a plurality of" is at least two, such as two, three or more, etc., unless otherwise explicitly specifically limited.

Claims

1. A method for detecting tension of yarn conveying, characterized in that: The following steps are involved: S1, collecting multimodal vibration data of the yarn in a preset detection area, the multimodal vibration data including a time-series image of the yarn vibration displacement obtained by a linear array camera and a vibration acoustic signal obtained by an acoustic sensor; S2, based on the yarn vibration displacement time series image, uses the optical flow method to calculate the velocity field of at least two spatially separated measurement points on the yarn, and combines the spectrum analysis results of the vibroacoustic signal to construct a spatiotemporal joint feature vector that characterizes the yarn vibration dynamics; S3, using the equipment systematic interference model associated with the yarn delivery speed, decouples the spatiotemporal joint eigenvector to separate the pure yarn vibration characteristics that eliminate the influence of equipment background vibration; S4, based on the vibration characteristics of pure yarn and combined with the preset constitutive parameters of the yarn, determines the instantaneous tension of the yarn by solving the inverse problem of the nonlinear wave equation coupled with tension, bending moment and air damping. The preset constitutive parameters include the yarn's linear density, bending stiffness and nonlinear elastic coefficient.

2. The yarn conveying tension detection method according to claim 1, characterized in that: In S1, a linear array camera is used to shoot at an acquisition frame rate that can capture the main vibration modes of the yarn; the acoustic sensor is placed in a preset detection area so that the acoustic sensor can pick up the vibration acoustic signal generated by the yarn vibration.

3. The yarn conveying tension detection method according to claim 2, characterized in that: In S1, the line array camera is arranged vertically facing the yarn.

4. The yarn conveying tension detection method according to claim 1, characterized in that: In S2, multiple measuring points distributed along the yarn conveying path are selected in the vibration displacement time series image, and the optical flow algorithm is used to calculate the vibration velocity time series of each measuring point perpendicular to the yarn conveying direction.

5. The yarn conveying tension detection method according to claim 3, characterized in that: In S2, the spectrum analysis of the vibroacoustic signal is performed to extract one or more main frequency features that characterize its main energy distribution; the main frequency features are fused with the vibration velocity time series features of multiple measuring points to form a spatiotemporal joint feature vector.

6. The yarn conveying tension detection method according to any one of claims 1 to 5, characterized in that: In S3, during the decoupling calculation, the equipment systematic interference model is constructed as a state-space model. The parameters of the state-space model are determined by calibrating the vibration characteristics of the equipment when it is running no-load at a specific conveying speed. A filtering-based state estimation algorithm is used, and the joint time-space feature vector is used as the observation input to estimate and separate the pure yarn vibration characteristics.

7. The yarn conveying tension detection method according to any one of claims 1 to 5, characterized in that: In S4, the nonlinear wave equation takes into account the bending stiffness of the yarn itself and the air damping effect during movement.

8. The yarn feeding tension detection method according to claim 7, characterized in that: In S4, an objective function is constructed with the instantaneous tension value of the yarn as the parameter to be optimized and the goal of minimizing the error between the vibration characteristics of the pure yarn and the vibration characteristics predicted by the nonlinear wave equation; an iterative optimization algorithm is used to solve the optimization problem and obtain the instantaneous tension value.

9. The yarn conveying tension detection method according to claim 8, characterized in that: The error between the pure yarn vibration characteristics and the vibration characteristics predicted by the nonlinear wave equation is the root mean square error.

10. A yarn conveying tension detection system, characterized in that: The method comprises a memory and a processor, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the method for detecting the tension of yarn conveying according to any one of claims 1 to 9 is implemented.

Images