Yarn breakage monitoring method

Abstract

The invention provides a yarn breakage monitoring method, which comprises the following steps of: controlling an executing mechanism of textile equipment according to a positioned position coordinate of a potential broken end point, automatically separating a yarn section of the potential broken end point, marking the yarn section, and sending position information of the potential broken end point and corresponding yarn section characteristic data to a monitoring center; and in the monitoring center, the received potential end breaking point information, the separated yarn section feature data and pre-stored historical end breaking data are subjected to correlation analysis, an Apriori correlation rule mining algorithm is adopted, the correlation rule between the potential end breaking point and the position where end breaking occurs is found, and the time and the position where end breaking occurs are pre-judged.

Description

Technical Field

[0001] The invention relates to the field of information technology, and in particular to a yarn breakage monitoring method. Background Art

[0002] Yarn breakage is a difficult and frequent problem in high-speed textile production. The vibration and tension changes generated by the yarn during high-speed operation are the main causes of yarn breakage. At present, most yarn breakage monitoring systems can only detect the occurrence of yarn breakage events, but cannot accurately locate the breakage position, nor can they warn of the risk of yarn breakage in real time.

[0003] Yarn breakage has an all-round impact on textile quality and production efficiency. Yarn breakage can lead to finished product defects, reduced yarn strength, and time-consuming repair of broken yarns, which seriously affects product quality and delivery time. The location of broken yarns is difficult to accurately locate, and manual search along the yarn is required, which results in low positioning efficiency.

[0004] An ideal yarn break monitoring solution should be able to sense the yarn status in real time, accurately locate potential break points, and issue an early warning before breakage occurs. This requires the monitoring device to collect yarn vibration, tension and other signals in real time, and to perform intelligent analysis on massive monitoring data. At the same time, the impact of yarn breakage has a certain delay, and product quality may not be significantly affected for a period of time after the breakage occurs. Therefore, monitoring the impact of breakage and product quality traceability is also a technical difficulty.

[0005] In addition, the implementation cost of the yarn break monitoring solution is also a practical problem. In textile production, the number of yarns is huge and the environment is complex and changeable, which places high demands on the performance and reliability of the monitoring device. The monitoring device needs to be able to adapt to the harsh environment of the workshop, such as high temperature, humidity, vibration, electromagnetic interference, etc., and can operate stably for a long time. At the same time, the cost of the monitoring device also needs to be controlled within a reasonable range to ensure the economy of the system.

[0006] In summary, a high-performance, low-cost, and easy-to-implement intelligent monitoring solution for yarn breakage is a technical problem that the textile industry needs to solve urgently. This requires the integration of advanced technologies such as intelligent sensing, big data analysis, and edge computing, and in-depth understanding of textile business scenarios to create innovative solutions. Summary of the invention

[0007] The present invention provides a yarn breakage monitoring method, which mainly comprises: Obtaining a yarn vibration signal and a tension signal, wherein the yarn vibration signal and the tension signal are acquired in real time by a plurality of vibration sensors and tension sensors arranged on the yarn, acquiring product batch information corresponding to the current yarn, and transmitting the acquired yarn vibration signal, tension signal and product batch information to an edge computing node; in the edge computing node, using a continuous wavelet transform algorithm to perform time-frequency domain analysis on the yarn vibration signal, extracting characteristic parameters reflecting the abnormal state of the yarn, wherein the characteristic parameters include energy entropy and singular value decomposition coefficients, performing statistical analysis on the yarn tension signal, and obtaining statistical characteristic quantities such as the mean, standard deviation, kurtosis and skewness of the tension; using a feature-level fusion method to integrate the wavelet time-frequency features of the yarn vibration signal and the statistical features of the yarn tension signal to obtain fusion features, inputting the fusion features into a pre-constructed support vector machine model, and evaluating the health status of the yarn. If the evaluation result exceeds a preset normal range threshold, it is determined that the section of yarn has a potential risk of breaking; For yarn segments that are judged to have a risk of breaking ends, an improved convolutional neural network algorithm is used to further analyze the vibration signal and tension signal of the yarn segment, and accurately locate the position coordinates of the potential break point, wherein the improved convolutional neural network adopts a residual connection structure to enhance the ability to extract multi-scale features; According to the position coordinates of the located potential break point, the actuator of the textile equipment is controlled to automatically separate the yarn segment of the potential break point and mark it, and the position information of the potential break point and the corresponding yarn segment feature data are sent to the monitoring center; In the monitoring center, the received potential break point information, the separated yarn segment feature data and the pre-stored historical break data are correlated and analyzed, and the Apriori association rule mining algorithm is used to find the correlation law between the potential break point and the position of the break that has occurred, and the time and position of the break are pre-determined; According to the predicted time and location of the end breakage, the yarn segment that may break is removed from the production line in advance, and the characteristic data of the yarn segment is associated with the previously recorded product batch information for subsequent product quality traceability. The predicted information is fed back to the production scheduling system to dynamically optimize the production plan and reduce the impact of the end breakage event on production efficiency.

[0008] The technical solution provided by the embodiment of the present invention may have the following beneficial effects: The present invention discloses a method for predicting and preventing yarn breakage. Multiple sensors are arranged on the yarn to collect vibration and tension signals in real time, and combined with product batch information, time-frequency domain analysis and feature fusion are performed at the edge computing node. A support vector machine model is used to evaluate the health status of the yarn and determine the potential risk of breakage. For yarn segments with risks, an improved convolutional neural network is used to accurately locate the potential breakage points. The present invention automatically separates and marks risky yarn segments and sends relevant information to a monitoring center. The monitoring center predicts the time and location of the breakage through correlation analysis and data mining. Based on this, risky yarn segments are eliminated in advance, the production plan is optimized, and the intelligent prediction and active prevention of yarn breakage are realized, which effectively improves the efficiency of textile production and product quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 The present invention is a flow chart of a yarn breakage monitoring method.

[0010] Figure 2 It is a schematic diagram of a yarn breakage monitoring method of the present invention.

[0011] Figure 3 It is another schematic diagram of a yarn breakage monitoring method of the present invention. DETAILED DESCRIPTION

[0012] In order to further understand the content of the present invention, the present invention is described in detail in conjunction with the accompanying drawings and embodiments. The present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It is understood that the specific embodiments described herein are only used to explain the relevant inventions, rather than to limit the invention. It is also necessary to explain that, for ease of description, only the parts related to the invention are shown in the accompanying drawings.

[0013] like Figure 1-3 In this embodiment, a yarn breakage monitoring method may specifically include: S101. Obtain yarn vibration signal and tension signal, wherein the yarn vibration signal and tension signal are collected in real time by multiple vibration sensors and tension sensors arranged on the yarn; obtain product batch information corresponding to the current yarn; and transmit the collected yarn vibration signal, tension signal and product batch information to the edge computing node.

[0014] Multiple vibration sensors and tension sensors are arranged on the yarn to collect the vibration signal and tension signal of the yarn in real time. The product batch information to which the current yarn belongs is obtained and associated with the collected vibration signal and tension signal. The collected yarn vibration signal, tension signal and product batch information are transmitted to the edge computing node. In the edge computing node, the received yarn vibration signal and tension signal are preprocessed, including denoising, normalization and other operations, to obtain the preprocessed vibration signal and tension signal. According to the preprocessed vibration signal and tension signal, characteristic parameters reflecting the yarn quality, such as vibration frequency, amplitude, tension mean, tension variance, etc. are extracted. The extracted characteristic parameters are input into the pre-trained support vector machine model to classify the yarn quality and determine whether the yarn has quality defects. If the judgment result shows that the yarn has quality defects, the batch of yarn is marked as defective, and the defect information is associated with the product batch information and stored in the database; if the judgment result shows that the yarn quality is qualified, the batch of yarn is marked as qualified, and the qualified information is associated with the product batch information and stored in the database.

[0015] Specifically, a vibration sensor and a tension sensor are arranged every 10 meters on the yarn production line, and the sampling frequency of the sensor is set to 1kHz. The yarn currently being produced is batch number 20230530001, and the batch number is associated with the collected vibration signal and tension signal to form a data packet. The data packet is transmitted to the edge computing server 20 meters away from the production line via industrial Ethernet. The edge computing server uses a Qualcomm Snapdragon 888 processor with a main frequency of up to 84GHz and runs a Linux operating system optimized for yarn quality detection. The received data packet is parsed to extract the vibration signal and tension signal, and then the signal is denoised by wavelet transform, and the signal amplitude is normalized to between 0 and 1 by the maximum and minimum normalization method. The frequency domain characteristics of the vibration signal are extracted using short-time Fourier transform, including the amplitude values ​​at frequencies of 10Hz, 20Hz, and 50Hz, as well as the frequency spectrum entropy, and the time domain characteristics of the tension signal are extracted, including the mean, variance, peak-to-peak value, etc. The 10 extracted feature parameters are input into the pre-trained support vector machine model. Using the radial basis kernel function, the model determines that the comprehensive score of the current yarn quality is 85, which exceeds the preset qualified threshold of 8, so the batch of yarn is determined to be qualified. The batch number 20230530001 is associated with the qualified mark and stored in the MySQL database of the edge computing server. The database is automatically synchronized to the cloud server every day to form a complete product quality traceability chain.

[0016] S102. In the edge computing node, a continuous wavelet transform algorithm is used to perform time-frequency domain analysis on the yarn vibration signal to extract characteristic parameters reflecting the abnormal state of the yarn, wherein the characteristic parameters include energy entropy and singular value decomposition coefficients; a statistical analysis is performed on the yarn tension signal to obtain statistical characteristic quantities such as the mean, standard deviation, kurtosis and skewness of the tension.

[0017] In the edge computing node, the yarn vibration signal and the yarn tension signal are obtained, and the yarn vibration signal and the yarn tension signal are preprocessed to remove noise interference. For the preprocessed yarn vibration signal, the continuous wavelet transform algorithm is used to perform time-frequency domain analysis to extract characteristic parameters such as energy entropy and singular value decomposition coefficients. According to the extracted energy entropy and singular value decomposition coefficients, a support vector machine model is constructed, and the model parameters are trained and optimized to obtain a classification model for identifying abnormal yarn conditions. For the preprocessed yarn tension signal, statistical analysis is performed to calculate statistical characteristic quantities such as tension mean, standard deviation, kurtosis and skewness. The calculated tension statistical characteristic quantity is compared with the preset normal yarn tension characteristic threshold to determine whether the yarn tension is abnormal. If it exceeds the threshold range, it is determined to be abnormal tension. The classification result of the yarn vibration signal is integrated with the abnormal judgment result of the yarn tension to comprehensively determine whether the yarn state is abnormal. If both are abnormal, the yarn state is determined to be abnormal. According to the abnormal judgment results of the yarn status, an alarm signal is triggered to notify the textile equipment to make corresponding adjustments or maintenance operations to ensure the stability of the textile production process and product quality.

[0018] Specifically, in the edge computing node, the yarn vibration signal and tension signal are collected by high-precision vibration sensors and tension sensors respectively, with a sampling frequency of 1000Hz. The collected signals are preprocessed, and a 5th-order Butterworth low-pass filter is used with a cutoff frequency set to 200Hz to effectively remove high-frequency noise interference. For the preprocessed yarn vibration signal, the db4 wavelet basis function is selected to perform a 5-layer continuous wavelet transform decomposition on the signal, and the energy entropy characteristics of each subband and the first three singular value decomposition coefficients are extracted to form a 9-dimensional feature vector. The support vector machine model is trained using 500 groups of normal yarn vibration signal and 500 groups of abnormal yarn vibration signal sample data. The radial basis kernel function is used to optimize the model parameters through grid search, and the cross-validation accuracy reaches more than 95%. For the preprocessed yarn tension signal, the sliding window method is used with a window length of 1000 data points to calculate the four statistical feature quantities of the tension signal in the window, including the mean, standard deviation, kurtosis and skewness. By analyzing 100 sets of normal yarn tension data, the tension characteristic threshold range under normal working conditions is obtained, with a mean of 10N~20N, a standard deviation of less than 2N, a kurtosis of 5~5, and a skewness of -2~2. The statistical characteristic quantity of yarn tension calculated in real time is compared with the threshold. If it exceeds the normal range, it is judged as abnormal tension. Finally, the classification result of the yarn vibration signal and the result of the tension abnormality judgment are logically "AND" operated to obtain the final comprehensive judgment result of the yarn state. When the yarn state is abnormal, the sound and light alarm is triggered immediately, and the abnormal information is uploaded to the data center to notify the textile equipment control system to adjust the tension or replace the yarn in time to avoid product quality problems. Through real-time signal processing and intelligent analysis of edge computing nodes, online monitoring of yarn state and abnormal warning are realized to ensure the stability of the textile production process and product quality.

[0019] S103. Using a feature-level fusion method, the wavelet time-frequency features of the yarn vibration signal and the statistical features of the yarn tension signal are integrated to obtain fused features; the fused features are input into a pre-constructed support vector machine model to evaluate the health status of the yarn. If the evaluation result exceeds a preset normal range threshold, it is determined that the yarn segment has a potential risk of breaking.

[0020] The vibration signal and tension signal of the yarn during the production process are collected, and the collected signal data are transmitted to the data processing unit. The yarn vibration signal is subjected to wavelet transform to extract its time-frequency domain features and obtain the wavelet time-frequency feature vector. The yarn tension signal is subjected to statistical analysis to extract its statistical features such as mean, variance, peak-to-peak value, and obtain the statistical feature vector. The wavelet time-frequency feature vector and the statistical feature vector are fused at the feature level to obtain the fused feature vector as the representation of the yarn health status. The fused feature vector is input into the pre-trained support vector machine model to evaluate the health status of the yarn and obtain the health status evaluation value. The health status evaluation value is compared with the preset normal range threshold. If the evaluation value exceeds the normal range, it is judged that the section of yarn has a potential risk of breaking. According to the result of the risk of breaking, an early warning signal is generated to notify the textile equipment to make corresponding adjustments or maintenance to ensure the yarn quality and production efficiency.

[0021] Specifically, during the yarn production process, the vibration signal and tension signal of the yarn are collected by the vibration sensor and the tension sensor respectively, with a sampling frequency of 1000Hz and a sampling time of 10 seconds for each sample. The collected signal data is transmitted to the data processing unit via industrial Ethernet. For the yarn vibration signal, the db4 wavelet basis function is used to perform a 5-layer wavelet decomposition, and the wavelet coefficients at each scale are extracted. The time-frequency domain features such as the mean, variance, and energy of the wavelet coefficients at each scale are calculated to form a 12-dimensional wavelet time-frequency feature vector. For the yarn tension signal, eight statistical features such as the mean, variance, peak-to-peak value, and root mean square are calculated to form an 8-dimensional statistical feature vector. The wavelet time-frequency feature vector and the statistical feature vector are cascaded to obtain a 20-dimensional fusion feature vector as a comprehensive representation of the yarn health status. Using 100 groups of normal yarn data and 100 groups of abnormal yarn data collected in advance, their fusion feature vectors are extracted, and the support vector machine model with radial basis kernel function is used for training to obtain a yarn health status assessment model. The currently collected yarn fusion feature vector is input into the model to obtain the health status assessment value, which ranges from 0 to 1. The larger the value, the healthier the yarn state. Through the analysis of historical data, the health status assessment value range of normal yarn is set to 7 to 0. If the current health status assessment value of the yarn is lower than 7, it is judged that there is a potential risk of yarn breakage in this section, and an early warning signal is generated to notify the textile equipment to adjust the tension or replace the yarn to ensure the yarn quality and production efficiency. Through this intelligent monitoring method, the health status of the yarn can be evaluated in real time, potential quality problems can be discovered in time, the occurrence of yarn breakage can be reduced, and the textile production efficiency can be improved.

[0022] S104. For yarn segments judged to be at risk of breakage, an improved convolutional neural network algorithm is used to further analyze the vibration signal and tension signal of the yarn segment to accurately locate the position coordinates of the potential breakage point, wherein the improved convolutional neural network adopts a residual connection structure to enhance the ability to extract multi-scale features.

[0023] The vibration signal and tension signal data of the yarn segment to be detected are obtained to construct a time series data set. The time series data is preprocessed, including denoising, normalization and other operations to improve data quality. According to the preprocessed time series data, multi-scale features are extracted, including time domain features, frequency domain features, etc. An improved convolutional neural network model is adopted, and a residual connection structure is introduced to enhance the extraction and fusion capabilities of multi-scale features. The extracted multi-scale features are input into the improved convolutional neural network model, and a breakage risk prediction model is obtained through training. The trained breakage risk prediction model is used to predict the vibration signal and tension signal of the yarn segment to be detected to determine whether there is a breakage risk. If there is a breakage risk, the abnormal fluctuations of the vibration signal and tension signal are further analyzed, and the position coordinates of the potential breakage point are accurately determined through the positioning algorithm.

[0024] Specifically, the vibration signal and tension signal data of the yarn segment are first collected by high-precision sensors with a sampling frequency of 1000Hz and a duration of 10 seconds to construct a time series data set containing 10,000 data points. The collected time series data are subjected to wavelet denoising, and the db4 wavelet basis function is selected with a decomposition layer of 5 to effectively remove high-frequency noise interference. Then, the denoised data is normalized to the maximum and minimum values, and the data is mapped to the [0,1] interval to eliminate the influence of the data dimension. Based on the normalized time series data, 10 time domain features such as mean, variance, peak-to-peak value, and 5 frequency domain features such as spectral entropy and center frequency are extracted to form a 15-dimensional multi-scale feature vector. An improved convolutional neural network model is designed, and residual connections are introduced between traditional convolutional layers. Feature reuse and fusion are achieved through cross-layer connections to enhance the feature extraction capability of the network. The model contains 3 convolutional layers and 2 fully connected layers. The convolution kernel sizes are 3×3, 5×5 and 7×7, respectively, and the corresponding receptive fields are 30ms, 50ms and 70ms, which can capture local features of different scales. The extracted 15-dimensional multi-scale features are input into the improved convolutional neural network model. The Adam optimizer is used with an initial learning rate of 0.01 and a batch size of 32. After training for 500 epochs, a breakage risk prediction model is obtained. The vibration signal and tension signal of the yarn segment to be tested are predicted using the trained model. If the output probability is greater than 8, it is determined that there is a breakage risk. The fluctuation of the abnormal signal is further analyzed, and the time-frequency energy distribution of the signal is calculated by short-time Fourier transform to find the time point of energy mutation. Combined with the transmission speed of the yarn segment, the spatial position coordinates of the potential breakage point are converted, and the positioning accuracy can reach 1cm.

[0025] S105. According to the located position coordinates of the potential break point, the actuator of the textile equipment is controlled to automatically separate the yarn segment of the potential break point and mark it; the position information of the potential break point and the corresponding yarn segment feature data are sent to the monitoring center.

[0026] The position coordinate data of the located potential breakpoint is obtained, and the specific position of the potential breakpoint in the textile equipment is determined through the coordinate mapping relationship. According to the position of the potential breakpoint, the actuator of the textile equipment is controlled to move to the position, and the yarn segment of the potential breakpoint is separated from the adjacent yarn by mechanical clamping or adsorption. The separated yarn segment of the potential breakpoint is coded and marked, and a unique identification number is generated, which is associated with the position coordinates, image features and other information of the yarn segment and stored. The image of the separated yarn segment of the potential breakpoint is extracted, and the characteristic parameters such as the diameter, cross-sectional morphology, and fiber arrangement of the yarn segment are analyzed by the image processing algorithm. The data such as the position coordinates, unique identification number, and image features of the potential breakpoint are packaged and sent to the monitoring center server through the communication module. After receiving the potential breakpoint data, the monitoring center stores it in the breakpoint database, triggers the breakpoint analysis program, and comprehensively analyzes the distribution law and causes of the potential breakpoint. According to the characteristic parameters and distribution of potential breakage points, the risk level of the breakage points is judged through machine learning algorithms such as support vector machines or decision trees, and early warning information is generated and fed back to the textile equipment to dynamically adjust the equipment parameters and reduce the probability of breakage.

[0027] Specifically, the three-dimensional coordinate data of potential break points are obtained through a high-precision laser rangefinder with an accuracy of up to 1mm. The coordinates are mapped to the mechanical coordinate system of the textile equipment using affine transformation, and the positioning accuracy reaches 5mm. The control system drives the robot arm to move to the target position according to the mapped coordinates. The robot adopts pneumatic grippers with adjustable clamping force to achieve accurate separation of the yarn without damaging the fiber. The separated yarn segments are printed with a unique 18-bit alphanumeric mixed code by a micro-injection printer, and are associated with parameters such as coordinates and shooting angles and stored in the break point database. The high-definition CCD camera is combined with a telecentric lens to obtain the surface morphology image of the yarn segment, and the contour is extracted by the Canny edge detection algorithm. The yarn diameter change curve is calculated by wavelet analysis, and the regularity of the fiber arrangement is analyzed by the grayscale co-occurrence matrix to obtain the key characteristic parameters reflecting the yarn quality. After receiving the break point data, the monitoring center server uses the improved Apriori association rule mining algorithm to analyze the distribution law of the break points in different parts and time periods of the equipment, and infers the cause of the breakage in combination with the yarn production process parameters. A breakage risk prediction model was built based on the random forest algorithm, and the model hyperparameters were optimized through grid search, with a prediction accuracy of over 95%. When a high-risk breakage point is identified, the monitoring center sends an early warning message to the textile equipment via the OPC UA protocol, and the equipment automatically reduces the draft ratio and adjusts the tension to reduce the occurrence of breakage. Through this intelligent breakage prevention system, the breakage rate of textile equipment can be reduced by 50% and production efficiency can be increased by more than 20%.

[0028] S106. In the monitoring center, the received potential break point information, the separated yarn segment feature data and the pre-stored historical break data are subjected to correlation analysis, and the Apriori association rule mining algorithm is used to find the correlation law between the potential break point and the position where the break has occurred, so as to prejudge the time and position of the break.

[0029] According to the potential break point information received by the monitoring center, the position coordinate data of the potential break point is obtained; according to the position coordinate data of the potential break point, the yarn segment feature data corresponding to the position of the potential break point is obtained from the pre-stored yarn segment feature data; according to the obtained yarn segment feature data, a similarity calculation algorithm is used to search for the historical break data with the highest similarity to the current yarn segment feature data in the pre-stored historical break data; according to the searched historical break data, the break position coordinates corresponding to the historical break data are obtained; according to the position coordinate data of the potential break point and the historical break position coordinates, the Apriori association rule mining algorithm is used to calculate the correlation between the potential break point and the historical break position; if the calculated correlation is greater than the preset correlation threshold, the potential break point is determined as a predicted break point; according to the position coordinates of the predicted break point and the break time in the historical break data, a time series prediction algorithm is used to predict the time when the predicted break point will break.

[0030] Specifically, after the monitoring center receives the potential break point information, the system automatically extracts the XYZ three-dimensional coordinates of the potential break point. Then, in the pre-stored yarn segment feature database, the spatial index is used to quickly locate the yarn segment corresponding to the coordinate, and 10 characteristic values ​​such as fiber diameter and tensile strength are extracted. Next, the improved cosine similarity algorithm is used to calculate the similarity with the 50,000 records in the historical break database one by one, and the 20 historical break data with the highest similarity are screened out. Using the break position coordinates of these 20 data, the Apriori algorithm is used to mine the 5 association rules with the highest confidence. If the number of rules with a confidence of more than 80% is not less than 3, the potential break point is determined to be a predicted break point. Finally, the break time of these 20 historical data is selected to construct an LSTM time series prediction model to predict the probability of the predicted break point breaking within the next hour. If the probability exceeds 90%, an early warning is immediately issued to the workshop management system to notify the workers to handle it in advance to avoid break accidents.

[0031] S107. According to the predicted time and location of the end breakage, the yarn segment that may break is removed from the production line in advance, and the characteristic data of the yarn segment is associated with the previously recorded product batch information for subsequent product quality traceability; the predicted information is fed back to the production scheduling system to dynamically optimize the production plan and reduce the impact of the end breakage event on production efficiency.

[0032] The running status of the yarn on the production line is monitored in real time through machine vision technology, and the image data and position information of the yarn are obtained to determine whether there is a risk of end breakage. Based on historical data and current monitoring data, time series prediction algorithms such as ARIMA model are used to predict the time and location of possible end breakage. If it is predicted that a certain yarn segment may break, the production line is controlled to automatically remove the yarn segment to avoid the impact of end breakage on subsequent production. The characteristic data of the removed yarn segment, such as yarn type, production time, production equipment, etc., are obtained and stored in association with the previously recorded product batch information to form a quality traceability chain. The predicted end breakage information is fed back to the production scheduling system, and the production plan is dynamically adjusted through optimization algorithms such as genetic algorithms, such as adjusting the production line speed, raw material delivery, etc., to minimize the impact of end breakage events. Various data in the production process are continuously collected, including equipment operating parameters, environmental factors, etc., and machine learning algorithms such as support vector machines are used to continuously optimize the accuracy of the end breakage prediction model. The data in the quality traceability chain is regularly analyzed to explore the key factors leading to end breakage, and feedback is given to the production management department to continuously improve the production process and equipment maintenance, and improve overall production efficiency and product quality.

[0033] Specifically, a high-speed camera is used to capture yarn images at a speed of 100 frames per second, and the yarn contour is extracted through an image segmentation algorithm. The change rate of the yarn diameter is calculated, and a change rate of more than 5% is considered to be a risk of breaking. The ARIMA (1,1,0) model is established using the breakage data of the past three months to predict the yarn segments with a probability of breaking exceeding 80% in the next hour. For the yarn segments that may break, they are removed by a robot, and the yarn type, production batch, landing time and other information are recorded and stored in the MySQL database. The yarn conveying speed is dynamically adjusted, and the speed is reduced by 20% when the predicted probability of breaking exceeds 60%. Ten parameters such as equipment vibration, temperature and humidity are collected every 30 minutes, and the breakage prediction model is established using the SVM algorithm with an accuracy rate of more than 95%. Association rule mining is performed on the traceability data, and it is found that the moisture content of the yarn is the main cause of the breakage. Feedback is given to the process department to optimize the drying process, and the breakage rate is reduced by 15%.

[0034] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A yarn breakage monitoring method, characterized in that: The method comprises: Obtaining a yarn vibration signal and a tension signal, wherein the yarn vibration signal and the tension signal are acquired in real time by a plurality of vibration sensors and tension sensors arranged on the yarn, acquiring product batch information corresponding to the current yarn, and transmitting the acquired yarn vibration signal, tension signal and product batch information to an edge computing node; In the edge computing node, using a continuous wavelet transform algorithm to perform time-frequency domain analysis on the yarn vibration signal, extracting characteristic parameters reflecting the abnormal state of the yarn, wherein the characteristic parameters include energy entropy and singular value decomposition coefficients, performing statistical analysis on the yarn tension signal, and obtaining statistical characteristic quantities such as the mean, standard deviation, kurtosis and skewness of the tension; Using a feature-level fusion method, integrating the wavelet time-frequency features of the yarn vibration signal and the statistical features of the yarn tension signal to obtain fusion features, inputting the fusion features into a pre-constructed support vector machine model, and evaluating the health status of the yarn. If the evaluation result exceeds a preset normal range threshold, it is determined that the yarn segment has a potential risk of breaking; For yarn segments that are judged to have a risk of breaking ends, an improved convolutional neural network algorithm is used to further analyze the vibration signal and tension signal of the yarn segment, and accurately locate the position coordinates of the potential break point, wherein the improved convolutional neural network adopts a residual connection structure to enhance the ability to extract multi-scale features; According to the position coordinates of the located potential break point, the actuator of the textile equipment is controlled to automatically separate the yarn segment of the potential break point and mark it, and the position information of the potential break point and the corresponding yarn segment feature data are sent to the monitoring center; In the monitoring center, the received potential break point information, the separated yarn segment feature data and the pre-stored historical break data are correlated and analyzed, and the Apriori association rule mining algorithm is used to find the correlation law between the potential break point and the position of the break that has occurred, and the time and position of the break are pre-determined; According to the predicted time and location of the end breakage, the yarn segment that may break is removed from the production line in advance, and the characteristic data of the yarn segment is associated with the previously recorded product batch information for subsequent product quality traceability. The predicted information is fed back to the production scheduling system to dynamically optimize the production plan and reduce the impact of the end breakage event on production efficiency.

2. A yarn breakage monitoring method according to claim 1, characterized in that: The method comprises acquiring a yarn vibration signal and a tension signal, wherein the yarn vibration signal and the tension signal are acquired in real time by a plurality of vibration sensors and tension sensors arranged on the yarn; acquiring product batch information corresponding to the current yarn; and transmitting the acquired yarn vibration signal, tension signal and product batch information to an edge computing node, including: Multiple vibration sensors and tension sensors are arranged on the yarn to collect the vibration signal and tension signal of the yarn in real time; Obtain the product batch information to which the current yarn belongs, and associate it with the collected vibration signal and tension signal; The collected yarn vibration signal and tension signal as well as product batch information are transmitted to the edge computing node; In the edge computing node, the received yarn vibration signal and tension signal are preprocessed, including denoising and normalization operations, to obtain the preprocessed vibration signal and tension signal; According to the pre-processed vibration signal and tension signal, characteristic parameters reflecting yarn quality are extracted, such as vibration frequency, amplitude, tension mean and tension variance. The extracted feature parameters are input into a pre-trained support vector machine model to classify the yarn quality and determine whether the yarn has quality defects; If the judgment result shows that the yarn has quality defects, the batch of yarn is marked as defective, and the defect information is associated with the product batch information and stored in the database; If the judgment result shows that the yarn quality is qualified, the batch of yarn is marked as qualified, and the qualified information is associated with the product batch information and stored in the database.

3. A yarn breakage monitoring method according to claim 1, characterized in that: In the edge computing node, a continuous wavelet transform algorithm is used to perform time-frequency domain analysis on the yarn vibration signal to extract characteristic parameters reflecting the abnormal state of the yarn, wherein the characteristic parameters include energy entropy and singular value decomposition coefficients; a statistical analysis is performed on the yarn tension signal to obtain statistical characteristic quantities such as the mean, standard deviation, kurtosis and skewness of the tension, including: In the edge computing node, the yarn vibration signal and the yarn tension signal are obtained, and the yarn vibration signal and the yarn tension signal are preprocessed to remove noise interference; The continuous wavelet transform algorithm is used to analyze the pre-processed yarn vibration signal in the time and frequency domain, and characteristic parameters such as energy entropy and singular value decomposition coefficients are extracted. According to the extracted energy entropy and singular value decomposition coefficients, a support vector machine model is constructed, and the model parameters are trained and optimized to obtain a classification model for identifying abnormal yarn conditions; Statistical analysis is performed on the pre-processed yarn tension signal to calculate statistical characteristic quantities such as tension mean, standard deviation, kurtosis and skewness; The calculated tension statistical characteristic quantity is compared with the preset normal yarn tension characteristic threshold value to determine whether the yarn tension is abnormal; If it exceeds the threshold range, it is judged as abnormal tension; The classification results of the yarn vibration signal are integrated with the abnormal judgment results of the yarn tension to comprehensively judge whether the yarn state is abnormal; If both are abnormal, the yarn state is determined to be abnormal; According to the abnormal judgment results of the yarn status, an alarm signal is triggered to notify the textile equipment to make corresponding adjustments or maintenance operations to ensure the stability of the textile production process and product quality.

4. A yarn breakage monitoring method according to claim 1, characterized in that: The feature-level fusion method is adopted to integrate the wavelet time-frequency features of the yarn vibration signal and the statistical features of the yarn tension signal to obtain fusion features; the fusion features are input into a pre-built support vector machine model to evaluate the health status of the yarn. If the evaluation result exceeds a preset normal range threshold, it is determined that the section of yarn has a potential risk of breaking, including: Collect vibration signals and tension signals of the yarn during the production process, and transmit the collected signal data to the data processing unit; Perform wavelet transform on the yarn vibration signal, extract its time-frequency domain features, and obtain the wavelet time-frequency feature vector; Statistical analysis is performed on the yarn tension signal to extract its statistical features such as mean, variance and peak-to-peak value, and obtain the statistical feature vector; The wavelet time-frequency feature vector and the statistical feature vector are fused at feature level to obtain the fused feature vector as the representation of the yarn health status. The fused feature vector is input into a pre-trained support vector machine model to evaluate the health status of the yarn and obtain a health status evaluation value; The health status assessment value is compared with the preset normal range threshold. If the assessment value exceeds the normal range, it is judged that the yarn segment has a potential risk of breaking. Based on the results of the end-break risk assessment, an early warning signal is generated to notify the textile equipment to make corresponding adjustments or maintenance to ensure yarn quality and production efficiency.

5. A yarn breakage monitoring method according to claim 1, characterized in that: For the yarn segment judged to have the risk of breaking, the improved convolutional neural network algorithm is used to further analyze the vibration signal and tension signal of the yarn segment to accurately locate the position coordinates of the potential breaking point, wherein the improved convolutional neural network adopts a residual connection structure to enhance the ability to extract multi-scale features, including: Obtain the vibration signal and tension signal data of the yarn segment to be detected and construct a time series data set; Preprocess time series data, including denoising and normalization, to improve data quality; Extract multi-scale features based on the preprocessed time series data, including time domain features and frequency domain features; An improved convolutional neural network model is used, and a residual connection structure is introduced to enhance the ability to extract and fuse multi-scale features; The extracted multi-scale features are input into the improved convolutional neural network model, and the decapitation risk prediction model is obtained through training. The trained end-break risk prediction model is used to predict the vibration signal and tension signal of the yarn segment to be tested to determine whether there is an end-break risk. If there is a risk of breakage, the abnormal fluctuations of the vibration signal and the tension signal are further analyzed, and the position coordinates of the potential breakage point are accurately determined through the positioning algorithm.

6. A yarn breakage monitoring method according to claim 1, characterized in that: The method comprises: controlling the actuator of the textile equipment according to the position coordinates of the potential break point located, automatically separating the yarn segment of the potential break point and marking it; sending the position information of the potential break point and the corresponding yarn segment feature data to the monitoring center, including: Obtaining the position coordinate data of the located potential break point, and determining the specific position of the potential break point in the textile equipment through the coordinate mapping relationship; According to the position of the potential break point, the actuator of the textile equipment is controlled to move to the position, and the yarn segment at the potential break point is separated from the adjacent yarns by mechanical clamping or adsorption. Encoding and marking the separated potential break point yarn segment to generate a unique identification number, which is associated with the position coordinates and image features of the yarn segment and stored; Extract the image of the separated potential break point yarn segment, and analyze the characteristic parameters such as the diameter, cross-sectional morphology and fiber arrangement of the yarn segment through image processing algorithm; Packing data such as the location coordinates of potential break points, unique identification numbers, and image features of yarn segments, and sending them to the monitoring center server through the communication module; After receiving the potential breakpoint data, the monitoring center stores it in the breakpoint database and triggers the breakpoint analysis program to comprehensively analyze the distribution pattern and causes of the potential breakpoints; According to the characteristic parameters and distribution of potential breakage points, the risk level of the breakage points is judged through machine learning algorithms such as support vector machines or decision trees, and early warning information is generated and fed back to the textile equipment to dynamically adjust the equipment parameters and reduce the probability of breakage.

7. A yarn breakage monitoring method according to claim 1, characterized in that: In the monitoring center, the received potential break point information, the separated yarn segment feature data and the pre-stored historical break data are correlated and analyzed, and the Apriori association rule mining algorithm is used to find the correlation between the potential break point and the position of the broken end, and the time and position of the break are predicted, including: According to the potential breakpoint information received by the monitoring center, the location coordinate data of the potential breakpoint is obtained; According to the position coordinate data of the potential break point, the yarn segment characteristic data corresponding to the position of the potential break point is obtained from the pre-stored yarn segment characteristic data; According to the acquired yarn segment characteristic data, a similarity calculation algorithm is used to search the pre-stored historical end-break data for the historical end-break data with the highest similarity to the current yarn segment characteristic data; According to the searched historical end-break data, the end-break position coordinates corresponding to the historical end-break data are obtained; According to the location coordinate data of potential breakpoints and the coordinates of historical breakpoints, the Apriori association rule mining algorithm is used to calculate the correlation between potential breakpoints and historical breakpoints. If the calculated correlation is greater than a preset correlation threshold, the potential breakpoint is determined as a predicted breakpoint; According to the position coordinates of the predicted breaking point and the breaking time in the historical breaking data, the time series prediction algorithm is used to predict the time when the predicted breaking point will break.

8. A yarn breakage monitoring method according to claim 1, characterized in that: According to the predicted time and location of the breakage, the yarn segment that may breakage is removed from the production line in advance, and the characteristic data of the yarn segment is associated with the previously recorded product batch information for subsequent product quality traceability; the predicted information is fed back to the production scheduling system to dynamically optimize the production plan and reduce the impact of the breakage event on production efficiency, including: Use machine vision technology to monitor the yarn running status on the production line in real time, obtain the image data and position information of the yarn, and determine whether there is a risk of yarn breakage; Based on historical data and current monitoring data, time series forecasting algorithms such as the ARIMA model are used to predict when and where breakage may occur; If it is predicted that a certain yarn segment may break, the production line will be controlled to automatically remove the yarn segment to avoid the breakage affecting subsequent production; Obtain characteristic data of the rejected yarn segments, such as yarn type, production time and production equipment, and store them in association with previously recorded product batch information to form a quality traceability chain; Feedback the predicted breakage information to the production scheduling system, and dynamically adjust the production plan through optimization algorithms such as genetic algorithms, such as adjusting the production line speed and raw material input, to minimize the impact of the breakage event; Continuously collect various data during the production process, including equipment operating parameters and environmental factors, and use machine learning algorithms such as support vector machines to continuously optimize the accuracy of the break prediction model; Regularly analyze the data in the quality traceability chain to identify the key factors that lead to breakage, and provide feedback to the production management department to continuously improve production processes and equipment maintenance, thereby improving overall production efficiency and product quality.

Images