A method and system for monitoring yarn motion state

Abstract

The present invention relates to the technical field of yarn motion state monitoring, specifically a yarn motion state monitoring method and system. First, the deformation of the yarn during motion is monitored and converted into tension data, which is uploaded to the yarn motion state database in real time; the yarn motion video is captured by an industrial camera, the offset and angle from the preset trajectory are calculated, and it is determined whether the offset coefficient exceeds the threshold, and the offset abnormal point and related tension data are recorded and uploaded in time; the breakage of the yarn is monitored by composite signal analysis in combination with acoustic and vibration signals, and the breakage tension and composite signal at the abnormal moment are recorded. In addition, the data in the database is analyzed by a Gaussian mixture model to accurately obtain the current motion state of the yarn, providing an efficient and reliable monitoring means for the state monitoring of the yarn motion during the textile process.

Description

Technical Field

[0001] The present invention relates to the technical field of yarn motion state monitoring, and in particular to a yarn motion state monitoring method and system. Background Art

[0002] As an important production equipment in the knitting industry, the production efficiency and product quality of the circular knitting machine depend to a large extent on the movement state of the yarn. When the circular knitting machine is working, the yarn needs to be guided, pulled and fed into the knitting needles for knitting under high-speed and continuous conditions. In this process, whether the tension, movement trajectory and running state of the yarn are stable directly affects the density uniformity, grain integrity and overall quality of the knitted fabric. However, due to factors such as equipment vibration, process parameter adjustment, uneven yarn material or external interference, the yarn may have abnormal problems such as excessive tension, movement deviation or even breakage. If these problems are not discovered and handled in time, it will not only lead to an increase in the defective rate of fabrics, but may also cause equipment wear or even shutdown, seriously affecting production efficiency.

[0003] Traditional yarn motion monitoring technologies mostly rely on sensors to obtain single parameters, such as tension detection or break detection, and are unable to fully capture the multi-dimensional state information of yarn motion. In addition, traditional methods have limited sensitivity and accuracy in abnormality detection, and are difficult to adapt to the complex and changing textile production environment, especially under high-speed operation conditions, where subtle changes in yarns are often easily overlooked. Current yarn motion monitoring technology is dominated by supervised learning with labeled information. Although supervised machine learning methods have significantly improved accuracy compared to traditional sensor-based monitoring technologies, a large amount of data annotation is required during the supervised learning process, and the real-time responsiveness of the monitoring stage is poor.

[0004] Therefore, a yarn motion state monitoring method and system are proposed. Summary of the invention

[0005] The purpose of the present invention is to provide a yarn motion state monitoring method and system, which monitors yarn deformation in real time and calculates yarn tension, and uploads it to a yarn motion state database, thereby ensuring real-time recording and tracking of tension data; at the same time, an industrial camera is used to obtain the yarn motion trajectory, and the offset and offset angle are calculated by comparing with the preset trajectory, dynamically generating the offset coefficient, accurately identifying the offset abnormal point, and recording the relevant abnormal information. Through the composite analysis of acoustic and vibration signals, real-time detection of break anomalies is further achieved, and the tension and composite signals at the time of break are recorded to provide data support for abnormal event analysis. Finally, the multimodal data in the database is modeled and analyzed through a Gaussian mixture model, and the motion state of the yarn is obtained in real time, providing a scientific and efficient solution for yarn motion state monitoring, abnormality detection and operation optimization in the textile process.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for monitoring yarn motion state, comprising:

[0008] Monitoring and acquiring yarn deformation generated during yarn movement at a first time interval, converting the yarn deformation into yarn tension, and uploading the yarn tension to a yarn movement state database;

[0009] Monitor and obtain the yarn motion trajectory, compare the yarn motion trajectory with the preset trajectory, obtain the offset and the offset angle; calculate and obtain the offset coefficient according to the offset and the offset angle; if the offset coefficient is greater than the offset threshold, mark the moment of yarn offset as the offset abnormal point, record the offset yarn tension and the offset abnormal coefficient at the offset abnormal point, and upload the offset yarn tension and the offset abnormal coefficient to the yarn motion state database; otherwise, obtain the offset coefficient at the first time interval and upload it to the yarn motion state database;

[0010] Monitor and obtain the acoustic signal and vibration signal generated when the yarn moves; calculate and obtain the composite signal according to the acoustic signal and the acoustic signal; mark the moment when the yarn breaks as the abnormal break point, record the broken yarn tension and the composite break signal at the abnormal break point, and upload the broken yarn tension and the composite break signal to the yarn movement state database; otherwise, obtain the composite signal at the first time interval and upload it to the yarn movement state database;

[0011] The data in the yarn motion state database is analyzed using a Gaussian mixture model to obtain the current yarn motion state.

[0012] Further, obtaining the yarn tension includes:

[0013] The light source emits an initial light signal, which is transmitted to the fiber Bragg grating area through the optical fiber and returns a backlight reflection signal; the backlight reflection signal is detected and the wavelength is compared with the initial light signal to obtain the Bragg wavelength change; the strain is calculated according to the Bragg wavelength change, and the yarn tension is calculated and obtained through the strain-tension elastic modeling formula.

[0014] Further, obtaining the offset coefficient includes:

[0015] An industrial camera is used to capture and obtain a video of yarn movement, and an image denoising algorithm is used to remove noise in the yarn movement video to obtain a denoised video; a background modeling algorithm is used to remove the static background of the denoised video, and the yarn contour is extracted. The yarn position is determined by calculating the center of mass of the yarn contour in each frame; the moving trajectory of the center of mass in each frame is tracked using an optical flow method, and the motion trajectory is smoothed using polynomial fitting to obtain a real-time trajectory of the yarn movement; the real-time trajectory of the yarn movement is compared with the preset trajectory, and the offset and the offset angle of each point are calculated; and the offset coefficient is calculated based on the offset and the offset angle.

[0016] Further, calculating and obtaining the composite signal includes:

[0017] Performing short-time Fourier transform on the acoustic signal and the vibration signal to obtain time-frequency spectrum of the acoustic signal and the vibration signal; calculating the real-time signal energy of the time-frequency spectrum of the acoustic signal and the vibration signal to obtain acoustic energy density and vibration energy density; performing energy-weighted fusion on the time-frequency spectrum of the acoustic signal and the vibration signal to obtain a composite time-frequency spectrum; performing inverse short-time Fourier transform on the composite time-frequency spectrum to obtain the composite signal.

[0018] Further, obtaining the yarn motion state includes:

[0019] The data in the yarn motion state database are standardized and missing value processed to obtain standardized data; the standardized data are serialized to obtain a serialized sample set; the serialized sample set is analyzed using a Gaussian mixture model and the model parameters are initialized; the Gaussian mixture model is trained using an expectation-maximization algorithm to optimize the model parameters; based on the trained Gaussian mixture model, the posterior logarithmic probability of the data at the current time point is calculated to obtain the current yarn motion state.

[0020] A yarn motion state monitoring system, comprising:

[0021] A yarn deformation monitoring module, used for monitoring and acquiring the yarn deformation generated when the yarn moves at a first time interval, converting the yarn deformation into yarn tension, and uploading the yarn tension to a yarn movement state database;

[0022] A yarn trajectory monitoring module is used to monitor and obtain the yarn motion trajectory, compare the yarn motion trajectory with a preset trajectory, obtain an offset and an offset angle; calculate and obtain an offset coefficient according to the offset and the offset angle; if the offset coefficient is greater than an offset threshold, mark the moment of yarn offset as an offset abnormal point, record the offset yarn tension and the offset abnormal coefficient at the offset abnormal point, and upload the offset yarn tension and the offset abnormal coefficient to the yarn motion state database; otherwise, obtain the offset coefficient at the first time interval and upload it to the yarn motion state database;

[0023] A yarn breakage monitoring module, used to monitor and obtain the acoustic signal and vibration signal generated when the yarn moves; calculate and obtain a composite signal according to the acoustic signal and the acoustic signal; mark the moment when the yarn breaks as a break abnormal point, record the broken yarn tension and the break composite signal at the break abnormal point, and upload the broken yarn tension and the break composite signal to the yarn movement state database; otherwise, obtain the composite signal at the first time interval and upload it to the yarn movement state database;

[0024] The real-time feedback module is used to analyze the data in the yarn motion state database using a Gaussian mixture model to obtain the current yarn motion state.

[0025] Further, obtaining the yarn tension includes:

[0026] The light source emits an initial light signal, which is transmitted to the fiber Bragg grating area through the optical fiber and returns a backlight reflection signal; the backlight reflection signal is detected and the wavelength is compared with the initial light signal to obtain the Bragg wavelength change; the strain is calculated according to the Bragg wavelength change, and the yarn tension is calculated and obtained through the strain-tension elastic modeling formula.

[0027] Further, obtaining the offset coefficient includes:

[0028] An industrial camera is used to capture and obtain a video of yarn movement, and an image denoising algorithm is used to remove noise in the yarn movement video to obtain a denoised video; a background modeling algorithm is used to remove the static background of the denoised video, and the yarn contour is extracted. The yarn position is determined by calculating the center of mass of the yarn contour in each frame; the moving trajectory of the center of mass in each frame is tracked using an optical flow method, and the motion trajectory is smoothed using polynomial fitting to obtain a real-time trajectory of the yarn movement; the real-time trajectory of the yarn movement is compared with the preset trajectory, and the offset and the offset angle of each point are calculated; and the offset coefficient is calculated based on the offset and the offset angle.

[0029] Further, calculating and obtaining the composite signal includes:

[0030] Performing short-time Fourier transform on the acoustic signal and the vibration signal to obtain time-frequency spectrum of the acoustic signal and the vibration signal; calculating the real-time signal energy of the time-frequency spectrum of the acoustic signal and the vibration signal to obtain acoustic energy density and vibration energy density; performing energy-weighted fusion on the time-frequency spectrum of the acoustic signal and the vibration signal to obtain a composite time-frequency spectrum; performing inverse short-time Fourier transform on the composite time-frequency spectrum to obtain the composite signal.

[0031] Further, obtaining the yarn motion state includes:

[0032] The data in the yarn motion state database are standardized and missing value processed to obtain standardized data; the standardized data are serialized to obtain a serialized sample set; the serialized sample set is analyzed using a Gaussian mixture model and the model parameters are initialized; the Gaussian mixture model is trained using an expectation-maximization algorithm to optimize the model parameters; based on the trained Gaussian mixture model, the posterior logarithmic probability of the data at the current time point is calculated to obtain the current yarn motion state.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] 1. The present invention effectively removes background noise in the video through a denoising algorithm, improves image quality, and ensures the accuracy of subsequent processing; extracts the yarn contour through a background modeling algorithm, and uses the optical flow method to track its center of mass, which can accurately obtain the motion trajectory of the yarn; polynomial fitting further smoothes the trajectory, eliminates noise interference, and makes the motion trajectory more stable and accurate. After comparing the actual motion trajectory with the preset trajectory, the yarn offset, offset angle, and offset coefficient are accurately calculated, thereby providing strong data support for real-time monitoring of the yarn motion state.

[0035] 2. The present invention obtains the time-frequency spectrum of the acoustic signal and the vibration signal respectively through short-time Fourier transform, and calculates their respective energy densities, which can reflect the frequency components and energy changes of the signals in real time; the time-frequency spectrum of the two is energy-weighted fused, which effectively combines the advantages of acoustic and vibration signals, enhances the perception of the yarn motion state, and is more sensitive in abnormal detection. The composite signal is restored by inverse short-time Fourier transform, and the fused information can be converted into a signal that can be used for actual monitoring, providing a more feasible and efficient means of monitoring the yarn motion state.

[0036] 3. This method realizes efficient identification and monitoring of yarn motion state by processing and modeling data in the yarn motion state database. By serializing the data by time, the time series information is retained, which helps to capture the dynamic changes of the yarn motion state over time; the Gaussian mixture model is used to model the serialized data, and the model parameters are optimized by the expectation maximization algorithm, so that the model can accurately fit the various modes of yarn motion state; finally, by calculating the posterior logarithmic probability of the data at each time point, the current yarn motion state can be judged in real time and accurately, providing a scientific basis and real-time feedback means for state monitoring and anomaly detection in the textile process. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 A flow chart of a yarn motion state monitoring method of the present invention;

[0038] Figure 2 The figure is a schematic diagram of the structure of a yarn motion state monitoring system of the present invention. DETAILED DESCRIPTION

[0039] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0040] Embodiment 1

[0041] A certain knitting factory A has multiple large circular knitting machines for producing daily knitted products. Due to the upgrade of mechanical equipment, the original yarn movement status monitoring method can no longer meet the current needs, resulting in frequent defective knitted products entering the market. The defects are mainly knitted surface damage caused by excessive tension, pattern deformation caused by yarn deviation, and knitted thread ends caused by yarn breakage. To solve this problem, the factory used a variety of yarn movement status monitoring methods on multiple large circular knitting machines for comparison to select the best solution. One of the large circular knitting machines adopted a yarn movement status monitoring method proposed in the present invention, such as Figure 1 As shown, including:

[0042] S100: monitoring and acquiring yarn deformation generated when the yarn moves at a first time interval, converting the yarn deformation into yarn tension, and uploading the yarn tension to a yarn movement state database;

[0043] Specifically, the first time interval may be 1 minute, 2 minutes, 5 minutes or longer. In the present embodiment, the first time interval is 1 minute.

[0044] Further, obtaining the yarn tension includes:

[0045] The light source emits an initial light signal, which is transmitted to the fiber Bragg grating area through the optical fiber and returns a backlight reflection signal; the backlight reflection signal is detected and the wavelength is compared with the initial light signal to obtain the Bragg wavelength change; the strain is calculated according to the Bragg wavelength change, and the yarn tension is calculated and obtained through the strain-tension elastic modeling formula.

[0046] Specifically, a fiber Bragg grating is a sensor commonly used to measure physical quantities such as strain and temperature. It periodically introduces changes in the refractive index in the optical fiber so that light of a specific wavelength is reflected. When the optical fiber is subjected to external forces (such as tension in the yarn), the periodic structure of the fiber Bragg grating changes, resulting in a shift in the wavelength of the reflected light. By measuring this wavelength change, strain or tension can be calculated.

[0047] In this embodiment, the optical fiber is laid along the yarn guide path to ensure that it keeps in contact with the yarn, and the fiber Bragg grating is installed in the middle section of the optical fiber; the light source of the optical fiber sends an initial light signal, which is transmitted to the fiber Bragg grating area through the optical fiber and returns a back light reflection signal, wherein the Bragg wavelength of the back light reflection signal is:

[0048] ;

[0049] in, represents the Bragg wavelength of the reflected light signal, n represents the effective refractive index of the optical fiber, Represents the period of the fiber Bragg grating; when the fiber is subjected to external strain, the physical length of the fiber and the period of the grating will change, resulting in a change in the reflected wavelength. The change in Bragg wavelength caused by strain can be expressed by the following formula:

[0050] ;

[0051] in, represents the change in Bragg wavelength, Indicates the strain-optical effect coefficient of the optical fiber, usually between 0.22 and 0.30. Indicates that the optical fiber is subjected to external strain; the strain is calculated according to the change in Bragg wavelength, and the yarn tension is obtained through the strain-tension elastic modeling formula, where the external strain The calculation formula is:

[0052] ;

[0053] The strain-tension elastic modeling formula is:

[0054] ;

[0055] Wherein, T represents the yarn tension, E represents the Young's modulus of the yarn, and A represents the cross-sectional area of ​​the yarn.

[0056] In this embodiment, high-precision real-time monitoring of yarn tension is achieved through fiber Bragg grating technology. Fiber Bragg grating can sense the slight deformation of the yarn and accurately obtain strain data through the wavelength change of the reflected light signal, and then calculate the yarn tension through the strain-tension elastic relationship model. This method has the advantages of non-contact, real-time and high sensitivity, can avoid the wear and interference of traditional mechanical sensors, and provide more stable and accurate yarn tension monitoring.

[0057] S200: Monitor and obtain the yarn motion trajectory, compare the yarn motion trajectory with the preset trajectory, obtain the offset and the offset angle; calculate and obtain the offset coefficient according to the offset and the offset angle; if the offset coefficient is greater than the offset threshold, mark the moment of yarn offset as the offset abnormal point, record the offset yarn tension and the offset abnormal coefficient at the offset abnormal point, and upload the offset yarn tension and the offset abnormal coefficient to the yarn motion state database; otherwise, obtain the offset coefficient at the first time interval and upload it to the yarn motion state database;

[0058] Specifically, the offset threshold is 0.5. When the offset coefficient is greater than the offset threshold, the time point is marked as an offset anomaly point, and the offset yarn tension and the offset anomaly coefficient are immediately recorded and uploaded to the yarn movement state database; otherwise, the offset coefficient is obtained at a first time interval and uploaded to the yarn movement state database, wherein the moment of obtaining the offset coefficient is synchronized with the moment of obtaining the yarn tension, that is, the yarn tension at the same moment can also be obtained while obtaining the offset coefficient.

[0059] Further, obtaining the offset coefficient includes:

[0060] An industrial camera is used to capture and obtain a video of yarn movement, and an image denoising algorithm is used to remove noise in the yarn movement video to obtain a denoised video; a background modeling algorithm is used to remove the static background of the denoised video, and the yarn contour is extracted. The yarn position is determined by calculating the center of mass of the yarn contour in each frame; the moving trajectory of the center of mass in each frame is tracked using an optical flow method, and the motion trajectory is smoothed using polynomial fitting to obtain a real-time trajectory of the yarn movement; the real-time trajectory of the yarn movement is compared with the preset trajectory, and the offset and the offset angle of each point are calculated; and the offset coefficient is calculated based on the offset and the offset angle.

[0061] Specifically, a high-definition industrial camera is installed on the yarn movement path. The camera needs to have a frame rate of more than 200Fps to shoot the yarn movement to capture the video of the yarn movement. The median filtering algorithm is used to remove the noise in the video to reduce background interference, and the histogram equalization enhancement method is used to improve the contrast, thereby obtaining the preprocessed video; the background modeling algorithm of the background subtraction method is used to remove the static background of the preprocessed video, the findContours() function of OpenCV is used to extract the yarn contour, and the moments() function of OpenCV is used to calculate the center of mass of each contour to obtain the position information of the yarn; the optical flow method is used to track the motion trajectory of the particle in each frame, the motion trajectory is smoothed using polynomial fitting, and the least squares method is used to solve the polynomial to obtain the real-time trajectory of the yarn movement; the real-time trajectory of the yarn movement is compared with the preset trajectory, and the offset and the offset angle of each point are calculated, wherein the preset trajectory is stored in the computer in advance by process design, and the calculation formula of the offset is:

[0062] ;

[0063] in, Indicates the offset, The horizontal coordinate representing the real-time trajectory of the yarn movement, Indicates the horizontal coordinate of the preset track, The vertical coordinate representing the real-time trajectory of the yarn movement, Indicates the vertical coordinate of the preset track. The calculation formula of the offset angle is:

[0064] ;

[0065] in, represents the offset angle, represents an inverse tangent function; the offset coefficient is obtained by calculating according to the offset amount and the offset angle, and the calculation formula of the offset coefficient is:

[0066] ;

[0067] in, Indicates the offset factor.

[0068] In this embodiment, by combining the background modeling algorithm, the optical flow method and the polynomial fitting technology, the precise tracking and offset monitoring of the yarn movement are achieved; by extracting the yarn contour and calculating the center of mass position, the motion trajectory of the yarn can be efficiently captured; the optical flow method and trajectory fitting technology are used to further improve the smoothness and accuracy of the motion trajectory; after comparing with the preset trajectory, the offset and offset angle of the yarn can be accurately calculated, thereby accurately evaluating the offset state of the yarn.

[0069] S300: performing short-time Fourier transform on the acoustic signal and the vibration signal to obtain a time-frequency spectrum of the acoustic signal and a time-frequency spectrum of the vibration signal; calculating the real-time signal energy of the time-frequency spectrum of the acoustic signal and the time-frequency spectrum of the vibration signal to obtain an acoustic energy density and a vibration energy density; performing energy-weighted fusion on the time-frequency spectrum of the acoustic signal and the time-frequency spectrum of the vibration signal to obtain a composite time-frequency spectrum; performing inverse short-time Fourier transform on the composite time-frequency spectrum to obtain the composite signal;

[0070] Specifically, the acoustic signal is acquired by an acoustic sensor installed near the yarn guide, and the vibration signal is acquired by a vibration sensor installed on the upper end of the tension disk; when the yarn breaks, the time point is marked as an abnormal deviation point, and the tension of the broken yarn and the break composite signal are immediately recorded and uploaded to the yarn motion state database; otherwise, the composite signal is acquired at a first time interval and uploaded to the yarn motion state database, wherein the moment of acquiring the composite signal is synchronized with the moment of acquiring the yarn tension, that is, the yarn tension at the same moment can also be acquired while acquiring the composite signal.

[0071] Further, calculating and obtaining the composite signal includes:

[0072] Performing short-time Fourier transform on the acoustic signal and the vibration signal to obtain time-frequency spectrum of the acoustic signal and time-frequency spectrum of the vibration signal; calculating the real-time signal energy of the time-frequency spectrum of the acoustic signal and the time-frequency spectrum of the vibration signal to obtain acoustic energy density and vibration energy density; performing energy-weighted fusion on the time-frequency spectrum of the acoustic signal and the time-frequency spectrum of the vibration signal to obtain a composite time-frequency spectrum; performing inverse short-time Fourier transform on the composite time-frequency spectrum to obtain the composite signal, wherein a partial acoustic signal and vibration signal and their corresponding composite signal data table are shown in Table 1.

[0073] Table 1. Data table of some acoustic signals and vibration signals and their corresponding composite signals

[0074]

[0075] Specifically, the acoustic signal and the vibration signal are subjected to short-time Fourier transform to obtain the acoustic signal time spectrum and the vibration signal time spectrum, wherein the calculation formulas of the acoustic signal time spectrum and the vibration signal time spectrum are respectively:

[0076] ;

[0077] Among them, t represents the time signal, f represents the frequency signal, represents the time-frequency spectrum of the acoustic signal, represents the acoustic signal acquired in real time, STFT() represents short-time Fourier transform, represents the time spectrum of the vibration signal, represents the vibration signal obtained in real time; the real-time signal energy calculated by calculating the spectrum of the acoustic signal and the spectrum of the vibration signal is obtained to obtain the acoustic energy density and the vibration energy density, wherein the calculation formulas of the acoustic energy density and the vibration energy density are respectively:

[0078] ;

[0079] in, represents the acoustic energy density, Represents the vibration energy density; the acoustic signal time spectrum and the vibration signal time spectrum are energy-weighted fused to obtain a composite time spectrum, where the calculation formula of the composite time spectrum is:

[0080] ;

[0081] in, represents the composite time-frequency spectrum, represents the acoustic signal weight, represents the weight of the vibration signal; the composite time-frequency spectrum is subjected to inverse short-time Fourier transform to obtain the composite signal, where the composite signal calculation formula is:

[0082] ;

[0083] in, represents a composite signal, represents the inverse short-time Fourier transform of the composite time-frequency spectrum.

[0084] In this embodiment, the changing characteristics of the signal in time and frequency can be obtained in real time through short-time Fourier transform, and the energy density of the calculated signal can further reflect the strength and changing trend of the signal; energy-weighted fusion allows the contribution of the two signals to be dynamically adjusted according to their relative importance, thereby improving the accuracy and reliability of the composite signal; finally, the composite signal is restored through inverse transformation, which not only retains the key information from the acoustic and vibration signals, but also more accurately reflects the abnormal changes caused by yarn breakage during yarn movement.

[0085] S400: Analyze the data in the yarn motion state database using a Gaussian mixture model to obtain the current yarn motion state.

[0086] Further, obtaining the yarn motion state includes:

[0087] The data in the yarn motion state database are standardized and missing value processed to obtain standardized data; the standardized data are serialized to obtain a serialized sample set; the serialized sample set is analyzed using a Gaussian mixture model and the model parameters are initialized; the Gaussian mixture model is trained using an expectation-maximization algorithm to optimize the model parameters; based on the trained Gaussian mixture model, the posterior logarithmic probability of the data at the current time point is calculated to obtain the current yarn motion state.

[0088] Table 2. Data in the yarn motion state database

[0089]

[0090] Specifically, as shown in Table 2, the data in the yarn motion state database are sorted in time series, and each data , Corresponding to the yarn tension, offset coefficient and composite signal respectively, the data in the yarn motion state database are standardized and the missing value is processed. The standardized calculation formula is:

[0091] ;

[0092] in represents standardized data, Represents the data in the database. represents the data mean, represents the data standard deviation, and then performs missing value processing on the standardized data, deleting samples with missing data; the standardized data is serialized by time to obtain a serialized sample set ; Select the Gaussian mixture model to analyze the serialized sample set and initialize the model parameters, where the model parameters include the mixing coefficient, mean vector and covariance matrix. The formula of the Gaussian mixture model is:

[0093] ;

[0094] in, Representation data The total probability density of , K represents the type of yarn motion state of data X, including normal operation, excessive tension, yarn deviation and yarn breakage, k represents the yarn motion state index of data X, represents the mixing coefficient of the k-th Gaussian distribution, represents the mean vector of the kth Gaussian distribution, represents the covariance matrix of the k-th Gaussian distribution, Represents the probability density of the kth Gaussian distribution at data X; the Gaussian mixture model is trained by the expectation maximization (EM) algorithm to optimize the model parameters. The EM algorithm includes the E step and the M step. The E step calculates the probability that each data point belongs to a Gaussian component. The calculation formula is:

[0095] ;

[0096] in, Representation data The probability of belonging to the kth Gaussian component. In the M step, according to the Gaussian component probability value calculated in the E step, the model parameters are updated to maximize the likelihood function of the data, where the model parameter update formula is:

[0097] ;

[0098] in, represents the updated mixing coefficient, represents the updated mean vector, Updated covariance matrix; According to the trained model, the posterior logarithmic probability of the data at each time point is calculated to determine the yarn motion state. The trained model can be expressed as:

[0099] ;

[0100] in, Represents data at a certain point in time The posterior log probability of .

[0101] In this embodiment, a Gaussian mixture model is used to model the serialized sample set, which can capture various motion states in the yarn motion data; the expectation maximization algorithm optimizes the model parameters so that the model can accurately fit the data; by calculating the logarithmic posterior probability of each time point, the motion state of the current yarn can be judged in real time, and abnormal states can be discovered in time, thereby improving the recognition accuracy and efficiency of abnormal yarn motion states.

[0102] A yarn motion state monitoring method used in this embodiment is compared with other monitoring methods, wherein circular knitting machine 1 adopts a traditional monitoring method, circular knitting machine 2 adopts an unsupervised KNN monitoring method, circular knitting machine 3 adopts a supervised learning-based SVM method, and circular knitting machine 4 adopts the monitoring method proposed by the present invention, to verify the effectiveness of the method in this embodiment. The comparative experimental results are shown in Table 3.

[0103] Table 3. Comparison of results obtained by various yarn motion state monitoring methods

[0104]

[0105] According to Table 3, the performance of the yarn motion state monitoring method proposed by the present invention is significantly improved compared with other monitoring methods. Among them, the detection rate of excessive tension is 95%, the detection rate of yarn deviation is 93%, and the detection rate of yarn breakage is 93%, which are 7%, 10% and 12% higher than the SVM-based monitoring method respectively. The average response speed of the yarn motion state monitoring method proposed by the present invention is 0.23 seconds, which is slightly lower than the traditional monitoring method, but it can effectively monitor the three abnormal states, and the detection rates of the three abnormal states all reach the optimal level.

[0106] This method obtains tension by monitoring yarn deformation, calculates the deviation coefficient by combining trajectory comparison to identify deviation anomalies, and simultaneously collects acoustic and vibration signals to generate composite signals to detect break anomalies. All data are uploaded to the database. The data is analyzed using a Gaussian mixture model to obtain the yarn motion state in real time, realize multi-dimensional monitoring and intelligent diagnosis, and improve the accuracy and efficiency of yarn motion state monitoring.

[0107] Embodiment 2

[0108] In the textile production line of knitting factory B, a yarn motion status monitoring system is applied. The system is equipped with a variety of monitoring modules, which can monitor and respond to yarn tension, yarn trajectory and yarn breakage in real time. The structure of the system is as follows Figure 2 As shown, including:

[0109] A yarn deformation monitoring module, used for monitoring and acquiring the yarn deformation generated when the yarn moves at a first time interval, converting the yarn deformation into yarn tension, and uploading the yarn tension to a yarn movement state database;

[0110] A yarn trajectory monitoring module is used to monitor and obtain the yarn motion trajectory, compare the yarn motion trajectory with a preset trajectory, obtain an offset and an offset angle; calculate and obtain an offset coefficient according to the offset and the offset angle; if the offset coefficient is greater than an offset threshold, mark the moment of yarn offset as an offset abnormal point, record the offset yarn tension and the offset abnormal coefficient at the offset abnormal point, and upload the offset yarn tension and the offset abnormal coefficient to the yarn motion state database; otherwise, obtain the offset coefficient at the first time interval and upload it to the yarn motion state database;

[0111] A yarn breakage monitoring module, used to monitor and obtain the acoustic signal and vibration signal generated when the yarn moves; calculate and obtain a composite signal according to the acoustic signal and the acoustic signal; mark the moment when the yarn breaks as a break abnormal point, record the broken yarn tension and the break composite signal at the break abnormal point, and upload the broken yarn tension and the break composite signal to the yarn movement state database; otherwise, obtain the composite signal at the first time interval and upload it to the yarn movement state database;

[0112] The real-time feedback module is used to analyze the data in the yarn motion state database using a Gaussian mixture model to obtain the current yarn motion state.

[0113] Further, obtaining the yarn tension includes:

[0114] The light source emits an initial light signal, which is transmitted to the fiber Bragg grating area through the optical fiber and returns a backlight reflection signal; the backlight reflection signal is detected and the wavelength is compared with the initial light signal to obtain the Bragg wavelength change; the strain is calculated according to the Bragg wavelength change, and the yarn tension is calculated and obtained through the strain-tension elastic modeling formula.

[0115] Further, obtaining the offset coefficient includes:

[0116] An industrial camera is used to capture and obtain a video of yarn movement, and an image denoising algorithm is used to remove noise in the yarn movement video to obtain a denoised video; a background modeling algorithm is used to remove the static background of the denoised video, and the yarn contour is extracted. The yarn position is determined by calculating the center of mass of the yarn contour in each frame; the moving trajectory of the center of mass in each frame is tracked using an optical flow method, and the motion trajectory is smoothed using polynomial fitting to obtain a real-time trajectory of the yarn movement; the real-time trajectory of the yarn movement is compared with the preset trajectory, and the offset and the offset angle of each point are calculated; and the offset coefficient is calculated based on the offset and the offset angle.

[0117] Further, calculating and obtaining the composite signal includes:

[0118] Performing short-time Fourier transform on the acoustic signal and the vibration signal to obtain time-frequency spectrum of the acoustic signal and the vibration signal; calculating the real-time signal energy of the time-frequency spectrum of the acoustic signal and the vibration signal to obtain acoustic energy density and vibration energy density; performing energy-weighted fusion on the time-frequency spectrum of the acoustic signal and the vibration signal to obtain a composite time-frequency spectrum; performing inverse short-time Fourier transform on the composite time-frequency spectrum to obtain the composite signal.

[0119] Further, obtaining the yarn motion state includes:

[0120] The data in the yarn motion state database are standardized and missing value processed to obtain standardized data; the standardized data are serialized to obtain a serialized sample set; the serialized sample set is analyzed using a Gaussian mixture model and the model parameters are initialized; the Gaussian mixture model is trained using an expectation-maximization algorithm to optimize the model parameters; based on the trained Gaussian mixture model, the posterior logarithmic probability of the data at the current time point is calculated to obtain the current yarn motion state.

[0121] As shown in Table 4, knitting factory B uses a yarn movement state monitoring system proposed in the present invention, which can effectively detect abnormal movement states generated during the yarn movement process, including excessive tension, yarn breakage, and yarn deviation. It can effectively reduce the defective rate of fabrics, reduce equipment wear and even downtime, and improve production efficiency.

[0122] Table 4. Data of some abnormal motion states and their detection results

[0123]

[0124] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for monitoring the motion state of a yarn, comprising: Monitoring and acquiring yarn deformation generated during yarn movement at a first time interval, converting the yarn deformation into yarn tension, and uploading the yarn tension to a yarn movement state database; Monitor and obtain the yarn motion trajectory, compare the yarn motion trajectory with the preset trajectory, obtain the offset and the offset angle; calculate and obtain the offset coefficient according to the offset and the offset angle; if the offset coefficient is greater than the offset threshold, mark the moment of yarn offset as the offset abnormal point, record the offset yarn tension and the offset abnormal coefficient at the offset abnormal point, and upload the offset yarn tension and the offset abnormal coefficient to the yarn motion state database; otherwise, obtain the offset coefficient at the first time interval and upload it to the yarn motion state database; Monitor and obtain the acoustic signal and vibration signal generated when the yarn moves; calculate and obtain the composite signal according to the acoustic signal and the acoustic signal; mark the moment when the yarn breaks as the abnormal break point, record the broken yarn tension and the composite break signal at the abnormal break point, and upload the broken yarn tension and the composite break signal to the yarn movement state database; otherwise, obtain the composite signal at the first time interval and upload it to the yarn movement state database; Using a Gaussian mixture model to analyze the data in the yarn motion state database to obtain the current yarn motion state; Acquiring the yarn motion state includes: standardizing and processing missing values ​​of data in the yarn motion state database to obtain standardized data; serializing the standardized data to obtain a serialized sample set; using a Gaussian mixture model to analyze the serialized sample set and initialize model parameters; using an expectation-maximization algorithm to train the Gaussian mixture model and optimize the model parameters; and according to the trained Gaussian mixture model, calculating the posterior logarithmic probability of the data at the current time point to obtain the current yarn motion state.

2. A yarn motion state monitoring method according to claim 1, characterized in that: Obtaining the yarn tension comprises: The light source emits an initial light signal, which is transmitted to the fiber Bragg grating area through the optical fiber and returns a backlight reflection signal; the backlight reflection signal is detected and the wavelength is compared with the initial light signal to obtain the Bragg wavelength change; the strain is calculated according to the Bragg wavelength change, and the yarn tension is calculated and obtained through the strain-tension elastic modeling formula.

3. A yarn motion state monitoring method according to claim 1, characterized in that: Obtaining the offset coefficient includes: An industrial camera is used to capture and obtain a video of yarn movement, and an image denoising algorithm is used to remove noise in the yarn movement video to obtain a denoised video; a background modeling algorithm is used to remove the static background of the denoised video, and the yarn contour is extracted. The yarn position is determined by calculating the center of mass of the yarn contour in each frame; the moving trajectory of the center of mass in each frame is tracked using an optical flow method, and the motion trajectory is smoothed using polynomial fitting to obtain a real-time trajectory of the yarn movement; the real-time trajectory of the yarn movement is compared with the preset trajectory, and the offset and the offset angle of each point are calculated; and the offset coefficient is calculated based on the offset and the offset angle.

4. A yarn motion state monitoring method according to claim 1, characterized in that: Calculating and obtaining the composite signal includes: Performing short-time Fourier transform on the acoustic signal and the vibration signal to obtain time-frequency spectrum of the acoustic signal and the vibration signal; calculating the real-time signal energy of the time-frequency spectrum of the acoustic signal and the vibration signal to obtain acoustic energy density and vibration energy density; performing energy-weighted fusion on the time-frequency spectrum of the acoustic signal and the vibration signal to obtain a composite time-frequency spectrum; performing inverse short-time Fourier transform on the composite time-frequency spectrum to obtain the composite signal.

5. A yarn motion state monitoring system, characterized in that: include: A yarn deformation monitoring module, used for monitoring and acquiring the yarn deformation generated when the yarn moves at a first time interval, converting the yarn deformation into yarn tension, and uploading the yarn tension to a yarn movement state database; A yarn trajectory monitoring module is used to monitor and obtain the yarn motion trajectory, compare the yarn motion trajectory with a preset trajectory, obtain an offset and an offset angle; calculate and obtain an offset coefficient according to the offset and the offset angle; if the offset coefficient is greater than an offset threshold, mark the moment of yarn offset as an offset abnormal point, record the offset yarn tension and the offset abnormal coefficient at the offset abnormal point, and upload the offset yarn tension and the offset abnormal coefficient to the yarn motion state database; otherwise, obtain the offset coefficient at the first time interval and upload it to the yarn motion state database; A yarn breakage monitoring module, used to monitor and obtain the acoustic signal and vibration signal generated when the yarn moves; calculate and obtain a composite signal according to the acoustic signal and the acoustic signal; mark the moment when the yarn breaks as a break abnormal point, record the broken yarn tension and the break composite signal at the break abnormal point, and upload the broken yarn tension and the break composite signal to the yarn movement state database; otherwise, obtain the composite signal at the first time interval and upload it to the yarn movement state database; A real-time feedback module is used to analyze the data in the yarn motion state database using a Gaussian mixture model to obtain the current yarn motion state; Acquiring the yarn motion state includes: standardizing and processing missing values ​​of data in the yarn motion state database to obtain standardized data; serializing the standardized data to obtain a serialized sample set; using a Gaussian mixture model to analyze the serialized sample set and initialize model parameters; using an expectation-maximization algorithm to train the Gaussian mixture model and optimize the model parameters; and according to the trained Gaussian mixture model, calculating the posterior logarithmic probability of the data at the current time point to obtain the current yarn motion state.

6. A yarn motion state monitoring system according to claim 5, characterized in that: Obtaining the yarn tension comprises: The light source emits an initial light signal, which is transmitted to the fiber Bragg grating area through the optical fiber and returns a backlight reflection signal; the backlight reflection signal is detected and the wavelength is compared with the initial light signal to obtain the Bragg wavelength change; the strain is calculated according to the Bragg wavelength change, and the yarn tension is calculated and obtained through the strain-tension elastic modeling formula.

7. A yarn motion state monitoring system according to claim 5, characterized in that: Obtaining the offset coefficient includes: An industrial camera is used to capture and obtain a video of yarn movement, and an image denoising algorithm is used to remove noise in the yarn movement video to obtain a denoised video; a background modeling algorithm is used to remove the static background of the denoised video, and the yarn contour is extracted. The yarn position is determined by calculating the center of mass of the yarn contour in each frame; the moving trajectory of the center of mass in each frame is tracked using an optical flow method, and the motion trajectory is smoothed using polynomial fitting to obtain a real-time trajectory of the yarn movement; the real-time trajectory of the yarn movement is compared with the preset trajectory, and the offset and the offset angle of each point are calculated; and the offset coefficient is calculated based on the offset and the offset angle.

8. A yarn motion state monitoring system according to claim 5, characterized in that: Calculating and obtaining the composite signal includes: Performing short-time Fourier transform on the acoustic signal and the vibration signal to obtain time-frequency spectrum of the acoustic signal and the vibration signal; calculating the real-time signal energy of the time-frequency spectrum of the acoustic signal and the vibration signal to obtain acoustic energy density and vibration energy density; performing energy-weighted fusion on the time-frequency spectrum of the acoustic signal and the vibration signal to obtain a composite time-frequency spectrum; performing inverse short-time Fourier transform on the composite time-frequency spectrum to obtain the composite signal.

Images