A yarn anti-winding method

Abstract

The application provides a yarn anti-winding method, comprising the following steps: according to the correlation between yarn material characteristics and processing technology, yarn raw material ratio and process parameter setting are optimized to reduce the generation of yarn surface hairiness impurities, improve yarn uniformity and smoothness, and thus reduce the risk of friction and adhesion between yarns; an automatic quality detection device is arranged between each process of yarn production, multi-source quality data such as yarn appearance and mechanics are collected, correlation analysis of yarn quality data is realized through Kalman filtering algorithm, a yarn quality traceability chain is constructed, and when yarn winding problems are found, the problem source can be quickly located and diagnosed to provide a basis for subsequent process adjustment.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of information technology, and particularly relates to a yarn anti-winding method. BACKGROUND

[0002] During the production process of yarn, due to the influence of factors such as material properties, processing technology, equipment state, etc., yarn winding problems are prone to occur. When the yarn is running at high speed, the surface hair, impurities, etc. are easy to cause friction and adhesion between the yarns, resulting in yarn winding together, forming quality defects such as yarn knots and broken ends. At the same time, when the yarn tension, speed and other parameters fluctuate, yarn winding problems are also easy to occur, affecting production efficiency and product quality.

[0003] In view of the technical problem of yarn winding, it is necessary to analyze and optimize from multiple aspects such as material properties, production process, equipment state, etc. First of all, the appropriate yarn raw material should be selected to reduce the surface hair and impurities of the yarn and improve the uniformity and smoothness of the yarn. Secondly, the production process of the yarn should be optimized, such as using air spinning, compact spinning and other advanced processes to reduce stress concentration and unevenness of the yarn in the production process. Thirdly, the maintenance and management of the yarn production equipment should be strengthened, such as regular cleaning, lubrication, calibration, etc. to ensure the stability and reliability of the equipment operation. Finally, the yarn quality monitoring and early warning mechanism should be established to monitor the key parameters of the yarn online, discover and deal with the quality problems such as yarn winding in time, and avoid the problem from being enlarged.

[0004] In summary, the yarn winding problem is a complex systematic problem, which needs to be optimized and controlled from multiple links such as yarn raw material, process, equipment, detection, etc. to effectively improve the yarn quality, reduce the occurrence probability of winding problem, and improve the production efficiency and economic benefit of textile enterprises. SUMMARY

[0005] The present application provides a yarn anti-winding method, which mainly comprises:

[0006] According to the relevance of yarn material properties and processing technology, by optimizing the yarn raw material ratio and process parameter setting, the generation of yarn surface hairiness impurities is reduced, the yarn uniformity and smoothness are improved, thereby reducing the risk of friction and adhesion between yarns, including adjusting fiber length distribution, fiber fineness, twist, sizing agent formula and other parameters;Using machine vision technology to analyze the real-time imaging of the yarn surface morphology, extracting key quality characteristic parameters such as the number and length distribution of yarn hairiness impurities, combining with the preset quality threshold, judging whether the yarn surface quality meets the standard in real time, when the yarn quality defect is detected, the alarm processing mechanism is automatically started, and the quality defect information is transmitted to the subsequent process;Install a vibration sensor on the yarn production equipment, collect the vibration signal of the equipment in real time, analyze the collected vibration signal in time and frequency domain, extract the abnormal vibration characteristics of the equipment, according to the preset equipment health threshold, judge whether the equipment running state is normal, when the equipment failure signs are detected, the equipment maintenance management process is automatically started, and the related process parameters are adjusted to reduce the influence on the yarn quality, at the same time, the equipment state information and the yarn quality detection result are associated, which is used to analyze the influence of equipment state on yarn quality;In the yarn production process, real-time collection of key process parameters such as yarn tension, speed, temperature and humidity, analysis of the correlation between yarn quality defects and process parameters by using decision tree algorithm, establishment of yarn quality prediction model based on random forest, when the process parameter fluctuation exceeds the preset range, timely warning signal is sent out, and production process is automatically adjusted to avoid yarn winding and other quality problems;Using laser speckle technology to precisely measure the surface morphology of the yarn, obtaining micro-morphology characteristic parameters such as yarn surface roughness and hairiness index, using support vector machine algorithm to construct yarn quality classification model, early warning of yarn winding tendency, according to the warning result, automatically optimizing the yarn production process, such as adjusting tension or speed, to reduce the risk of yarn winding, taking the laser speckle measurement result as one of the inputs of the process parameter prediction model, improving the prediction accuracy;Set up automatic quality detection device between each process of yarn production, collect multi-source quality data such as yarn appearance and mechanics, realize the correlation analysis of yarn quality data through Kalman filtering algorithm, construct yarn quality traceability chain, when yarn winding problem is found, quickly locate the problem source and diagnose, provide basis for subsequent process adjustment;Establish a yarn production whole-process quality control system, integrate multi-source heterogeneous data such as yarn raw material, process, equipment and quality, use principal component analysis and neural network algorithm to mine the complex correlation between yarn quality influencing factors, form yarn quality optimization decision scheme, according to the optimization scheme, automatically adjust the production parameters, continuously improve the yarn quality, improve the production efficiency and economic benefit.

[0007] The technical scheme provided by the embodiment of the application can include the following beneficial effects:

[0008] The application discloses a yarn production quality intelligent management and control method. The method reduces yarn surface hairiness impurities, improves uniformity and smoothness by optimizing raw material ratio and process parameters. Real-time analysis of yarn surface morphology is carried out by using machine vision technology, combined with vibration sensor monitoring equipment state, real-time collection of process parameters and establishment of quality prediction model. Laser speckle technology is used to measure yarn micro-morphology, and a quality classification model is constructed for early warning. Automatic detection devices are set between each process to realize quality data correlation analysis and traceability. The application integrates multi-source heterogeneous data, mines the complex correlation of quality influencing factors, forms an optimization decision scheme and automatically adjusts production parameters. Through the above technical means, the application realizes intelligent quality management and control of the whole process of yarn production, effectively reduces the occurrence of yarn winding and other quality problems, and improves product quality and production efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0009] Fig. 1 A flow chart of a yarn anti-winding method of the application.

[0010] Fig. 2 A schematic diagram of a yarn anti-winding method of the application.

[0011] Fig. 3 Another schematic diagram of a yarn anti-winding method of the application. DETAILED DESCRIPTION

[0012] For a further understanding of the present application, reference will be made to the following detailed description taken in conjunction with the accompanying drawings. The present application will be further described with reference to the drawings, in which:

[0013] As Figs. 1-3 , the yarn anti-winding method of the embodiment can specifically include:

[0014] Step S101, according to the correlation between yarn material properties and processing technology, the generation of yarn surface hairiness impurities is reduced, the uniformity and smoothness of the yarn are improved, and the friction and adhesion risk between the yarns is reduced by optimizing the yarn raw material ratio and process parameter setting. Specifically, it includes adjusting fiber length distribution, fiber fineness, twist, sizing agent formula and other parameters.

[0015] According to the yarn material properties, an optimal model of yarn raw material ratio is established, and the optimal raw material ratio scheme is determined by genetic algorithm to reduce the probability of yarn surface hairiness and impurities. Image processing technology is used to detect and analyze the yarn surface morphology, extract key parameters such as fiber length distribution and fineness, and combine with expert knowledge base to optimize the spinning process parameter settings using decision tree algorithm. Real-time data of yarn production process is collected by online sensors, and support vector machine algorithm is used to predict yarn uniformity and smoothness. When the predicted value is lower than the preset threshold, the twist and sizing agent formula are dynamically adjusted to ensure stable yarn quality. A yarn quality evaluation index system is constructed, considering factors such as hairiness index, impurity content, uniformity, etc. Fuzzy comprehensive evaluation method is used to quantitatively evaluate the yarn quality, providing decision basis for production optimization. Collecting yarn production data throughout the whole process, including raw material properties, process parameters, quality test results, etc. Based on big data analysis technology, key factors affecting yarn quality are mined to form a quality optimization knowledge base, guiding production practice. A yarn production expert system is developed, integrating intelligent modules such as raw material ratio, process optimization, quality prediction, etc. to realize intelligent control of the whole yarn production process, improve production efficiency and product quality stability. A yarn quality traceability system is established, realizing the whole life cycle tracking of raw materials, semi-finished products and finished products through Internet of Things technology. When quality problems occur, the problem can be quickly located to realize closed-loop management of quality problems and continuously optimize the production process.

[0016] Specifically, in order to optimize the yarn material ratio, the fiber length, fineness, strength and other attribute data of different materials can be collected, and a mathematical model of material attributes and yarn quality is established. Using genetic algorithm, the minimum hairiness index and impurity content are used as the optimization goal, the range of material ratio is set as the constraint condition, and the optimal material ratio scheme is obtained after 500 iterations, cross probability 8 and mutation probability 1. The cotton ratio is 60%, modal 30% and spandex 10%. The yarn surface morphology is imaged and collected, the resolution is 800x600, the gray scale is binarized, the fiber contour is extracted, the average fiber length is calculated as 15mm, and the fineness is 2dtex. The extracted attribute parameters are input into the decision tree model, and the spinning process parameters are set according to the expert knowledge base, including the speed of the spinning frame 800m / min, the ring speed 12000r / min, etc. The on-line sensor is installed on the spinning frame, the sampling frequency is 1kHz, and the yarn tension, hairiness index and other parameters are collected in real time. Using support vector machine algorithm, a quality prediction model is established, the training set has 2000 samples, the test set has 500 samples, and the prediction accuracy is 90%. When the yarn evenness prediction value is less than 85%, the twist is dynamically adjusted to 800tpm, the sizing agent ratio is PVA 5% and acrylic 3%, and the yarn strength and smoothness are improved. Based on the fuzzy comprehensive evaluation model of 5 indexes such as hairiness index, impurity content and strength, the membership function adopts Gaussian function, and the weight vector is determined by AHP method, and the yarn quality score is 85, which is in good level. A total of 100,000 yarn production process parameters and quality detection data are collected, and the association rule mining algorithm is used with support 0.5 and confidence 0.8 to find the positive correlation between the speed of the spinning frame and the yarn strength, and generate 50 quality optimization rules. A yarn production expert system based on B / S architecture is developed, the knowledge base covers 200 rules of material selection, process optimization and quality prediction, the inference engine adopts forward chain reasoning, and the whole process intelligent control is realized. By using RFID technology, electronic tags are attached to raw materials, semi-finished products and finished products, and quality traceability is realized through Internet of Things. When the yarn breakage rate exceeds 5%, the system automatically locates the process where the problem occurs, optimizes the spinning process parameters, and realizes closed-loop quality control.

[0017] In step S102, machine vision technology is used to analyze the surface morphology of the yarn in real time, extract key quality characteristic parameters such as the number and length distribution of hair and impurities, and combine with the preset quality threshold to judge whether the yarn surface quality meets the standard in real time. When the yarn quality defect is detected, the alarm processing mechanism is automatically started, and the quality defect information is transmitted to the subsequent process.

[0018] The yarn surface is continuously imaged by a high-resolution industrial camera to obtain high-definition digital images of the yarn surface. The obtained yarn surface images are pre-processed, including image denoising, enhancement and segmentation, etc., to improve the image quality and the accuracy of feature extraction. Image feature extraction algorithms such as texture analysis and edge detection are used to extract key quality feature parameters such as hairiness impurities and fiber ends from the pre-processed yarn surface images. Based on the extracted yarn surface quality feature parameters, a support vector machine is used to construct a yarn surface quality evaluation model to realize quantitative evaluation of the yarn surface quality. The yarn surface quality evaluation results are compared with the preset quality threshold value, and if the evaluation result is lower than the threshold value, it is determined as a quality defect, triggering an alarm processing mechanism. When a quality defect is detected, the defect type, location, severity, etc. information is transmitted in real time to the workshop management system and the subsequent process through the industrial bus communication protocol, so as to timely handle and quality trace. Big data analysis technology is used to statistically analyze the collected yarn surface quality data, to find the rules and influencing factors of quality defects, and to provide data support for optimizing spinning process parameters and improving equipment performance.

[0019] Specifically, an industrial camera with a resolution of 4096x3072 pixels is used to continuously image the yarn surface at a speed of 100 frames per second to obtain high-definition digital images of the yarn surface. The obtained yarn surface images are pre-processed, including median filter denoising, histogram equalization enhancement and threshold segmentation, etc., to improve the image quality and the accuracy of feature extraction. Gray level co-occurrence matrix texture analysis and Canny edge detection algorithms are used to extract 10 key quality feature parameters such as hairiness density, impurity area and fiber end number from the pre-processed yarn surface images. Based on the extracted yarn surface quality feature parameters, a support vector machine algorithm is used to construct a yarn surface quality evaluation model with a radial basis kernel function, and through 5-fold cross-validation, the model accuracy is above 95%, realizing quantitative evaluation of the yarn surface quality. The yarn surface quality evaluation results are compared with the preset quality threshold value, and if the evaluation result is lower than the threshold value of 8, it is determined as a quality defect, triggering an audible and light alarm processing mechanism. When a quality defect is detected, the defect type, location coordinates, severity, etc. information is transmitted in real time to the MES workshop management system and the subsequent drawing process through the EtherCAT industrial bus communication protocol, so as to timely handle and quality trace. Big data analysis technology is used to statistically analyze the collected yarn surface quality data, to find the correlation between yarn quality defects and spinning process parameters through correlation analysis and data mining, and to use a decision tree algorithm to construct a quality defect influencing factor analysis model, to provide data support and decision basis for optimizing spinning process parameters and improving equipment performance.

[0020] Step S103, install a vibration sensor on the yarn production equipment to collect real-time vibration signals of the equipment. Perform time-frequency domain analysis on the collected vibration signals to extract abnormal vibration features of the equipment. According to the preset equipment health threshold, determine whether the equipment running state is normal. When detecting equipment failure signs, automatically start the equipment maintenance management process and adjust related process parameters to reduce the impact on yarn quality. At the same time, associate the equipment state information with the yarn quality detection results for analyzing the influence of equipment state on yarn quality.

[0021] By installing a vibration sensor on the yarn production equipment, real-time vibration signal data of the equipment is collected. The collected vibration signal data is preprocessed, including denoising, normalization and other operations to improve data quality. Time-frequency domain analysis methods such as wavelet transform or Fourier transform are used to extract features from the preprocessed vibration signals to obtain abnormal vibration features of the equipment. According to the preset equipment health threshold, determine whether the extracted abnormal vibration features are outside the normal range. If it is outside the normal range, it is considered that the equipment running state is abnormal and there is a failure sign. When detecting equipment failure signs, automatically trigger the equipment maintenance management process by adjusting related process parameters such as speed, tension, etc. to reduce the impact of abnormal equipment state on yarn quality. Correlate and analyze the equipment state information with the yarn quality detection results, use machine learning algorithms such as support vector machine or decision tree to establish a prediction model between equipment state and yarn quality. Use the established prediction model to predict the yarn quality produced under the current equipment state in real time. When the predicted quality is lower than the standard, adjust the process parameters or maintain the equipment in time to ensure that the yarn production quality is stable within the qualified range.

[0022] Specifically, a piezoelectric vibration sensor of PCB 603C01 model is installed on the yarn production equipment, the sampling frequency is set to 5000 Hz, and the vibration data is transmitted to the data acquisition terminal in real time through the RS-485 bus. The collected vibration signal data is wavelet denoised, Daubechies4 wavelet basis function is selected for 5-layer decomposition of the signal, high-frequency noise components in the vibration signal are removed, and then Z-score normalization method is used to normalize the amplitude of the denoised signal, so that the signal amplitude is distributed in the interval of 0-1. Short-time Fourier transform is used for time-frequency analysis of the preprocessed vibration signal, the spectral energy value of the signal in the frequency band of 0-500 Hz is extracted as the abnormal vibration feature, and whether the current equipment is abnormal is judged by comparing the spectral energy threshold under the normal operation state of the equipment. When the equipment is detected to be abnormal, the equipment speed is automatically reduced by 10%, and the tension setting value is increased by 5N, so as to reduce the influence of the abnormal state of the equipment on the yarn quality. The device vibration characteristics and the yarn quality detection results are analyzed, the support vector machine regression algorithm is used to establish a prediction model of the device vibration energy value and the yarn strength, elongation at break and other quality indicators, the input of the model is the device vibration spectral energy value, and the output is the predicted value of the yarn quality indicators. When the quality prediction value is lower than the national standard requirement, the system automatically alarms and optimizes the adjustment of the equipment process parameters, and triggers the equipment maintenance management process to check and repair the equipment, so as to ensure the continuous and stable quality of the yarn.

[0023] In step S104, during the yarn production process, the key process parameters such as yarn tension, speed, temperature and humidity are collected in real time. The decision tree algorithm is used to analyze the correlation between yarn quality defects and process parameters, and a yarn quality prediction model based on random forest is established. When the process parameter fluctuation exceeds the preset range, a warning signal is sent in time, and the production process is automatically adjusted to avoid the occurrence of yarn winding and other quality problems.

[0024] Real-time data of key process parameters such as yarn tension, yarn speed, ambient temperature and ambient humidity in the yarn production process are acquired, and the data are transmitted to a data processing module. The data processing module pre-processes the acquired process parameter data, including data cleaning, data normalization and the like, to obtain a standardized process parameter data set. The standardized process parameter data set is input to a correlation rule analysis module based on a decision tree algorithm, and the correlation rules between yarn quality defects and each process parameter are mined through the decision tree algorithm to obtain a correlation rule set. According to the correlation rule set, a yarn quality prediction model is established in combination with a random forest algorithm, the standardized process parameter data set is taken as the input of the model, and the yarn quality prediction model is trained. In the yarn production process, each process parameter data is collected in real time, the data are input to the yarn quality prediction model, and the yarn quality prediction result under the current process parameters is obtained. It is judged whether the yarn quality prediction result exceeds a preset quality threshold range, if the threshold range is exceeded, a warning signal is triggered, and the warning signal is transmitted to a production control system. The production control system automatically adjusts the yarn production process parameters such as yarn tension, yarn speed and the like according to the warning signal and in combination with a preset process parameter adjustment strategy, so as to avoid the occurrence of yarn quality problems and ensure the yarn production quality.

[0025] Specifically, during the yarn production process, key process parameter data such as yarn tension, yarn speed, ambient temperature, and ambient humidity are collected in real time by sensors, with 10 collections per second. The collected data is transmitted to a data processing module via industrial Ethernet. The data processing module uses the Pandas library of Python to clean the acquired process parameter data, removing missing values and outliers. Then, the data is normalized using the min-max normalization method, mapping the data to the [0, 1] interval to obtain a standardized process parameter dataset. The standardized dataset is input into a correlation rule analysis module based on the C5 decision tree algorithm. By calculating the information gain ratio of each attribute, the attribute with the highest information gain ratio is selected as the root node of the decision tree, and the decision tree is recursively constructed to mine the association rules between yarn quality defects and process parameters, such as yarn tension exceeding 10 N and yarn speed exceeding 200 m / min, which can easily lead to yarn breakage. Based on the association rule set, combined with the random forest algorithm, a yarn quality prediction model is established using the sklearn library. The standardized process parameter dataset is divided into a training set and a test set in a ratio of 8:2. The model hyperparameters are optimized using the grid search and cross-validation methods, and the yarn quality prediction model is trained, with an accuracy of over 95%. During the yarn production process, process parameter data is collected every 5 seconds, and the data is input into the yarn quality prediction model to obtain the yarn quality prediction result under the current process parameters. If the prediction result exceeds the pre-set quality threshold range (such as yarn strength below 10 cN / tex), an early warning signal is triggered, and the early warning signal is transmitted to the production control system via the OPC UA protocol. The production control system adjusts the process parameters of the yarn production equipment automatically according to the early warning signal and pre-set process parameter adjustment strategies (such as reducing the yarn speed and increasing the yarn tension when the yarn strength is below the threshold). For example, the yarn speed is reduced from 200 m / min to 180 m / min, and the yarn tension is increased from 10 N to 12 N, to ensure that the yarn quality is stable within the acceptable range. Through the intelligent yarn quality prediction and control system, real-time monitoring and optimized control of the yarn production process can be achieved, effectively improving the yarn quality and production efficiency.

[0026] In step S105, the laser speckle technique is used to precisely measure the surface morphology of the yarn, obtaining micro-morphology characteristic parameters such as yarn surface roughness and fuzz index. A yarn quality classification model is constructed using the support vector machine algorithm to provide early warning of the yarn winding tendency. According to the warning result, the yarn production process is automatically optimized, such as adjusting the tension or speed, to reduce the risk of yarn winding. The laser speckle measurement result is used as one of the inputs of the process parameter prediction model, improving the prediction accuracy.

[0027] The yarn surface is scanned by a high-resolution laser speckle instrument to obtain three-dimensional topography data of the yarn surface. The speckle images obtained are processed to extract micro-topography characteristic parameters such as yarn surface roughness and hairiness index. The extracted topography characteristic parameters are analyzed according to the yarn quality standard to determine whether the yarn surface quality meets the requirements. If the yarn surface quality does not meet the requirements, the topography characteristic parameters are input into a pre-constructed support vector machine quality classification model to predict the yarn winding tendency. According to the prediction result of the support vector machine model, it is determined whether the yarn has a winding risk, and if there is a winding risk, early warning is triggered. When it is monitored that the yarn has a winding risk, the tension and drafting speed parameters of the yarn production equipment are automatically adjusted to optimize the production process and reduce the yarn winding probability. The yarn surface topography characteristic parameters measured by the laser speckle measurement are used together with the production equipment parameters as inputs of the process parameter prediction model, and the model is continuously optimized by machine learning algorithm to improve the accuracy of process parameter prediction.

[0028] Specifically, the yarn surface is scanned by a high-resolution laser speckle instrument with a resolution of 1 μm, and the scanning frequency is 1 kHz to obtain three-dimensional topography data of the yarn surface. The obtained speckle images are denoised by using a Gaussian filtering algorithm, and then the yarn surface profile is extracted by using an Otsu threshold segmentation algorithm. Micro-topography characteristic parameters such as yarn surface roughness Ra and hairiness index HI are extracted based on the profile. According to the yarn quality standard GB / T 398-2008, when the yarn surface roughness Ra is greater than 5 μm or the hairiness index HI is greater than 2, it is determined that the yarn surface quality is unqualified. The extracted topography characteristic parameters are input into a support vector machine quality classification model based on RBF kernel function, and the yarn winding tendency is predicted by a one-versus-all multi-classification algorithm. When the yarn winding probability is greater than 65%, it is determined that there is a winding risk and early warning is triggered. After it is monitored that the yarn has a winding risk, the tension and drafting speed parameters of the yarn production equipment are automatically adjusted based on a fuzzy PID control algorithm to optimize the production process by reducing the tension by 10% and increasing the drafting speed by 5% to reduce the yarn winding probability. The yarn surface roughness, hairiness index and other topography characteristic parameters measured by the laser speckle measurement are used together with the tension, drafting speed and other production equipment parameters as inputs of a BP neural network process parameter prediction model, and the model is trained by an error back propagation algorithm. When the prediction error is less than 5%, the iteration is stopped, and the model is continuously optimized to improve the accuracy of process parameter prediction.

[0029] Step S106, an automatic quality detection device is arranged between each process of yarn production to collect multi-source quality data such as yarn appearance and mechanics. The correlation analysis of yarn quality data is realized by Kalman filtering algorithm to construct a yarn quality traceability chain. When the yarn winding problem is found, the problem source is quickly located and diagnosed to provide basis for subsequent process adjustment.

[0030] According to the characteristics of each process of yarn production, the position of the quality detection point and the detection parameters are determined, the automatic quality detection device is installed at each detection point, and the appearance image data and mechanical property data of the yarn are collected in real time. The collected yarn appearance image data is pre-processed, the yarn surface defect features are extracted, and the defect recognition and classification are performed according to the preset defect type and threshold value, so as to obtain the yarn appearance quality evaluation result. The collected yarn mechanical property data is subjected to feature extraction and statistical analysis, and the quality evaluation is performed according to the preset mechanical property index and threshold value, so as to obtain the yarn mechanical quality evaluation result. The yarn appearance quality evaluation result and the mechanical quality evaluation result are fused, a multi-source quality data set is constructed, the Kalman filtering algorithm is used for correlation analysis and fusion of the multi-source quality data, and a comprehensive quality evaluation result is obtained. According to the yarn production process and the sequence of each process, the comprehensive quality evaluation result is arranged in time sequence, a yarn quality traceability chain is constructed, and the whole-process tracking and management of the yarn quality are realized. When the yarn winding problem is detected, the time node at which the problem occurs is traced back through the quality traceability chain, and the quality data and process parameters at the time node are used to perform root cause analysis by using a machine learning algorithm, so as to locate the key factors causing the yarn winding problem. According to the positioning result of the yarn winding problem, the process parameters of the corresponding process are optimized and adjusted, the adjusted process parameters are verified and evaluated, and the optimization is iterated until the yarn winding problem is eliminated and the yarn quality is stable and controllable.

[0031] Specifically, quality detection points are set in key processes such as spinning and winding in yarn production, high-speed cameras and tension sensors are installed to collect yarn surface images at a speed of 100 frames per second and yarn tension data at a frequency of 1 kHz. The Canny algorithm is used to detect the edges of the yarn image, extract the roundness and hairiness index of the yarn cross-section, and classify defects according to the gray level and shape difference of the yarn defects using the support vector machine (SVM) algorithm, with an identification rate of more than 95%. The frequency domain analysis of the yarn tension data is performed to extract the spectral entropy and harmonic distortion rate, and the threshold values of mechanical defects such as yarn breakage and neps are set to realize real-time evaluation of the mechanical properties of the yarn. The Kalman filtering algorithm is used to fuse the yarn surface quality data and mechanical property data, and the state equation and observation equation are established to realize dynamic tracking of the data through the prediction and update steps, with a fusion data accuracy rate of more than 98% at each quality detection point. The quality detection results of each batch of yarn are recorded in the manufacturing execution system (MES), and the yarn quality traceability chain is constructed through time stamp, batch number and other information to realize step-by-step tracing from the finished yarn to the raw material. When there is a yarn winding problem, the winding process is located through the quality traceability chain, the spinning process quality data of the batch is taken out, the decision tree algorithm is used to analyze the association rules between yarn winding and spinning quality parameters, and it is determined that the yarn hairiness content is the key factor causing winding. By adjusting the drafting ratio, airflow pressure and other process parameters of the spinning machine, the yarn hairiness index is optimized, and the quality data of the spinning process is monitored. After three iterations of optimization, the yarn winding rate is finally reduced to less than 5%, realizing stable control of yarn quality.

[0032] Step S107, a yarn production whole-process quality management and control system is established to integrate multi-source and heterogeneous data such as yarn raw materials, process, equipment and quality. Principal component analysis and neural network algorithms are used to mine the complex correlations between yarn quality influencing factors to form a yarn quality optimization decision scheme. According to the optimization scheme, production parameters are automatically adjusted to continuously improve yarn quality, production efficiency and economic benefits.

[0033] The multi-source heterogeneous data of raw materials, process, equipment and quality in the yarn production process are acquired, and the data are converted into structured data through data cleaning, conversion and integration. For the integrated structured data, the principal component analysis algorithm is used to extract the main influencing factors of yarn quality and reduce the data dimension. According to the results of principal component analysis, a neural network model is constructed to deeply mine the complex nonlinear relationship between the influencing factors and the yarn quality. Through the trained neural network model, the yarn quality under different production parameter combinations is predicted, and the optimal parameter combination is selected as the quality optimization scheme. The quality optimization scheme is converted into control instructions executable by the production equipment, and the production parameters are adjusted in real time through the automatic control system. The optimized production data are continuously collected, and the principal component analysis and neural network model are updated regularly to form a closed-loop control of continuous optimization of yarn quality. A yarn quality control visualization platform is established to monitor the key quality indicators of each production link in real time and provide data support for management decisions.

[0034] Specifically, first, the raw material parameters (such as cotton length, micronaire value, etc.), process parameters (such as carding process, spinning process, etc.), equipment parameters (such as spindle speed, draft ratio, etc.), and quality data (such as evenness, breaking strength, etc.) in the yarn production process are acquired through a data acquisition system. The multi-source heterogeneous data are cleaned, converted and integrated using ETL tools to form a structured data table. Then, the principal component analysis algorithm is used to reduce the dimension of the integrated structured data. By calculating the variance contribution rate of each variable, the main factors that have a greater impact on yarn quality are extracted, such as fiber length, spinning fineness, etc. On this basis, a three-layer BP neural network model is constructed, with the number of input layer nodes being the number of main influencing factors, the number of hidden layer nodes being 5 times the number of input layer nodes, and the number of output layer nodes being the number of yarn quality indicators. The learning rate is set to 0.1, and the iteration number is set to 1000 times to mine the nonlinear relationship between the influencing factors and the yarn quality. The trained neural network model is used to predict the yarn quality under different production parameters (such as spindle speed of 12000 rpm, draft ratio of 5, etc.), and the genetic algorithm is used to optimize the parameter combination to obtain the optimal quality scheme (such as spindle speed of 11500 rpm, draft ratio of 2, etc.). The optimization scheme is converted into PLC control instructions to adjust the parameters such as spindle speed and draft ratio of the spinning machine in real time. At the same time, the optimized production data are continuously collected, and the principal component analysis and neural network model are updated once a week to continuously improve the accuracy of yarn quality prediction. Finally, a Web-based yarn quality control platform is developed to display the trends of key quality indicators such as evenness and breaking strength in real time through line charts, and set quality warning thresholds to automatically alarm when the indicators exceed the thresholds, providing a basis for management decisions.

[0035] It is apparent that a person skilled in the art can make various modifications and variations to the embodiments of the application without departing from the spirit and scope of the application. Therefore, the application is intended to cover the modifications and variations of this application provided they come within the scope of the appended claims and their equivalents.

Claims

1. A yarn untwisting method characterized by, The method comprises: According to the correlation between yarn material properties and processing technology, by optimizing the yarn raw material ratio and process parameter setting, the generation of yarn surface hairiness impurities is reduced, the yarn uniformity and smoothness are improved, thereby reducing the risk of friction adhesion between yarns, specifically including adjusting the fiber length distribution, fiber fineness, twist, sizing agent formula parameters; Real-time imaging analysis of the yarn surface morphology is carried out by using machine vision technology, the number of yarn hairiness impurities, length distribution key quality characteristic parameters are extracted, combined with the preset quality threshold, the yarn surface quality is judged in real time whether it meets the standard, when the yarn quality defect is detected, the alarm processing mechanism is automatically started, and the quality defect information is transmitted to the subsequent process; Install a vibration sensor on the yarn production equipment, real-time collect the vibration signal of the equipment, analyze the collected vibration signal in time and frequency domain, extract the abnormal vibration characteristics of the equipment, according to the preset equipment health threshold, judge whether the equipment running state is normal, when the equipment failure signs are detected, the equipment maintenance management process is automatically started, and the related process parameters are adjusted to reduce the influence on the yarn quality, at the same time, the equipment state information and the yarn quality detection result are associated, which is used to analyze the influence of equipment state on yarn quality; In the yarn production process, the key process parameters of yarn tension, speed, temperature and humidity are collected in real time, the correlation between yarn quality defects and process parameters is analyzed by using decision tree algorithm, a yarn quality prediction model based on random forest is established, when the process parameter fluctuation exceeds the preset range, an early warning signal is sent in time, the production process is automatically adjusted to avoid the occurrence of yarn winding quality problems; The laser speckle technology is used to precisely measure the surface morphology of the yarn, the surface roughness, hairiness index and micro-morphology characteristic parameters of the yarn are obtained, a yarn quality classification model is constructed by using support vector machine algorithm, early warning of yarn winding tendency is carried out, according to the warning result, the yarn production process is automatically optimized, such as adjusting tension or speed, to reduce the risk of yarn winding, the laser speckle measurement result is used as one of the inputs of the process parameter prediction model, the prediction accuracy is improved; Set up automatic quality detection devices between each process of yarn production, collect multi-source quality data of yarn appearance and mechanics, realize the correlation analysis of yarn quality data by Kalman filtering algorithm, construct the yarn quality traceability chain, when the yarn winding problem is found, the problem source is quickly located and diagnosed, which provides basis for subsequent process adjustment; A yarn production whole-process quality management system is established, multi-source heterogeneous data of yarn raw materials, process, equipment and quality are integrated, principal component analysis and neural network algorithm are used to mine the complex correlation between yarn quality influencing factors, form a yarn quality optimization decision scheme, according to the optimization scheme, the production parameters are automatically adjusted, the yarn quality is continuously improved, the production efficiency and economic benefit are improved.

2. A yarn untwisting method according to claim 1, characterized in that, According to the correlation between yarn material properties and processing technology, by optimizing the yarn raw material ratio and process parameter setting, the generation of yarn surface hairiness and impurities is reduced, the uniformity and smoothness of the yarn are improved, and the risk of friction and adhesion between the yarns is reduced; Specifically including adjusting the fiber length distribution, fiber fineness, twist, sizing agent formula parameters, including: According to the yarn material properties, an optimized yarn raw material ratio model is established, and the optimal raw material ratio scheme is determined by genetic algorithm to reduce the probability of yarn surface hairiness and impurity generation; Image processing technology is used to detect and analyze the surface morphology of the yarn, extract the key parameters of fiber length distribution and fineness, combine with expert knowledge base, and use decision tree algorithm to optimize the spinning process parameter setting; Through real-time collection of yarn production process data by online sensors, support vector machine algorithm is used to predict the uniformity and smoothness of the yarn, and when the predicted value is lower than the preset threshold, the twist and sizing agent formula are dynamically adjusted to ensure the stability of the yarn quality; A yarn quality evaluation index system is constructed, considering the factors of hairiness index, impurity content and uniformity, and using fuzzy comprehensive evaluation method to quantitatively evaluate the yarn quality, providing decision basis for production optimization; Collect yarn production data throughout the whole process, including raw material properties, process parameters and quality test results, based on big data analysis technology, mine the key factors affecting yarn quality, form a quality optimization knowledge base, and guide production practice; Develop a yarn production expert system, integrate raw material ratio, process optimization and quality prediction intelligent modules, realize intelligent control of the whole process of yarn production, improve production efficiency and product quality stability; Establish a yarn quality traceability system, realize the whole life cycle tracking of raw materials, semi-finished products and finished products through Internet of Things technology, when quality problems occur, the problem can be quickly located, the closed-loop management of quality problems is realized, and the production process is continuously optimized.

3. A yarn untwisting method according to claim 1, wherein The machine vision technology is used to realize real-time imaging analysis of the surface morphology of the yarn, the key quality characteristic parameters of the yarn hairiness and impurity quantity and length distribution are extracted, and the preset quality threshold is combined to judge whether the surface quality of the yarn meets the standard in real time; When the yarn quality defect is detected, the alarm processing mechanism is automatically started, and the quality defect information is transmitted to the subsequent process, including: High-resolution industrial cameras are used to continuously image the surface of the yarn to obtain high-definition digital images of the yarn surface; The obtained yarn surface images are preprocessed, including image denoising, enhancement and segmentation, to improve the image quality and the accuracy of feature extraction; Image feature extraction algorithms such as texture analysis and edge detection are used to extract key quality characteristic parameters of hairiness and impurity and fiber breakage from the preprocessed yarn surface images; According to the extracted yarn surface quality characteristic parameters, support vector machine machine learning algorithm is used to construct yarn surface quality evaluation model to realize quantitative evaluation of yarn surface quality; The yarn surface quality evaluation result is compared with the preset quality threshold, if the evaluation result is lower than the threshold, it is judged as quality defect, and the alarm processing mechanism is triggered; When a quality defect is detected, the defect type, location, severity information is transmitted in real time to the workshop management system and subsequent processes through an industrial bus communication protocol, so as to timely handle and quality trace; Big data analysis technology is adopted to statistically analyze the collected yarn surface quality data, to mine the rules and influencing factors of quality defects, and to provide data support for optimizing spinning process parameters and improving equipment performance.

4. A yarn untwisting method according to claim 1, wherein The vibration sensor is installed on the yarn production equipment to collect the vibration signals of the equipment in real time; the collected vibration signals are analyzed in time and frequency domain to extract the abnormal vibration characteristics of the equipment; According to the preset equipment health threshold, it is judged whether the equipment running state is normal; when the equipment failure signs are detected, the equipment maintenance management process is automatically started, and the related process parameters are adjusted to reduce the influence on the yarn quality; at the same time, the equipment state information and the yarn quality detection results are associated to analyze the influence of the equipment state on the yarn quality, including: By installing vibration sensors on the yarn production equipment, real-time vibration signal data of the equipment is collected; The collected vibration signal data is preprocessed, including denoising and normalization operation to improve data quality; Time-frequency domain analysis methods such as wavelet transform or Fourier transform are used to extract features from preprocessed vibration signals to obtain abnormal vibration characteristics of the equipment; According to the preset equipment health threshold, it is judged whether the extracted abnormal vibration characteristics are out of the normal range, if so, it is considered that the equipment running state is abnormal, and there is a failure sign; When the equipment failure signs are detected, the equipment maintenance management process is automatically triggered, and the related process parameters such as speed and tension are adjusted to reduce the influence of the equipment abnormal state on the yarn quality; The equipment state information and the yarn quality detection results are associated and analyzed, and a prediction model between the equipment state and the yarn quality is established by using machine learning algorithms such as support vector machine or decision tree; The established prediction model is used to predict the yarn quality produced under the current state of the equipment in real time, and when the predicted quality is lower than the standard, the process parameters are adjusted or the equipment is maintained in time to ensure that the yarn production quality is stable within the qualified range.

5. A yarn untwisting method according to claim 1, wherein In the yarn production process, real-time yarn tension, speed, temperature and humidity key process parameters are collected; decision tree algorithm is used to analyze the correlation between yarn quality defects and process parameters, and a yarn quality prediction model based on random forest is established; when the process parameter fluctuation exceeds the preset range, a warning signal is sent in time to automatically adjust the production process to avoid yarn winding quality problems, including: Real-time data of yarn tension, yarn speed, environmental temperature and environmental humidity key process parameters in the yarn production process are obtained, and the data are transmitted to the data processing module; The data processing module preprocesses the obtained process parameter data, including data cleaning and data normalization operation, to obtain a standardized process parameter data set; The standardized process parameter data set is input into the association rule analysis module based on the decision tree algorithm, and the association rules between yarn quality defects and each process parameter are mined by the decision tree algorithm to obtain an association rule set; According to the association rule set, a yarn quality prediction model is established in combination with a random forest algorithm, a standardized process parameter data set is taken as an input of the model, and the yarn quality prediction model is trained; In the yarn production process, real-time acquisition of each process parameter data is performed, the data is input into the yarn quality prediction model, and a yarn quality prediction result under the current process parameter is obtained; It is judged whether the yarn quality prediction result exceeds a preset quality threshold range, if the threshold range is exceeded, a warning signal is triggered, and the warning signal is transmitted to a production control system; The production control system automatically adjusts the yarn production process parameters, such as adjusting the yarn tension and the yarn speed, according to the warning signal and in combination with a preset process parameter adjustment strategy, so as to avoid the occurrence of yarn quality problems and ensure the yarn production quality.

6. A yarn untwisting method according to claim 1, wherein The laser speckle technology is used to precisely measure the yarn surface morphology, to obtain the yarn surface roughness, the micro-morphology characteristic parameters of the hairiness index, and to construct a yarn quality classification model by using the support vector machine algorithm to perform early warning on the yarn winding tendency; According to the warning result, the yarn production process is automatically optimized, such as adjusting the tension or the speed, to reduce the yarn winding risk; the laser speckle measurement result is taken as one of the inputs of the process parameter prediction model to improve the prediction accuracy, including: A high-resolution laser speckle instrument is used to scan the yarn surface to obtain the three-dimensional morphology data of the yarn surface; The obtained speckle image is subjected to image processing to extract the yarn surface roughness and the micro-morphology characteristic parameters of the hairiness index; According to the yarn quality standard, the extracted morphology characteristic parameters are analyzed to determine whether the yarn surface quality meets the requirements; If the yarn surface quality does not meet the requirements, the morphology characteristic parameters are input into a pre-constructed support vector machine quality classification model to predict the yarn winding tendency; According to the prediction result of the support vector machine model, it is determined whether the yarn has a winding risk, if the yarn has a winding risk, early warning is triggered; When it is monitored that the yarn has a winding risk, the tension and the draft speed parameters of the yarn production equipment are automatically adjusted to optimize the production process and reduce the yarn winding probability; The yarn surface morphology characteristic parameters measured by the laser speckle are taken together with the production equipment parameters as the inputs of the process parameter prediction model, the model is continuously optimized by using the machine learning algorithm, and the accuracy of the process parameter prediction is improved.

7. A yarn untwisting method according to claim 1, wherein The automatic quality detection device is arranged between each process of the yarn production to collect the yarn appearance and mechanical multi-source quality data; the Kalman filtering algorithm is used to realize the correlation analysis of the yarn quality data to construct a yarn quality traceability chain; when the yarn winding problem is found, the problem source is quickly located and diagnosed to provide a basis for subsequent process adjustment, including: According to the characteristics of each process of the yarn production, the positions of the quality detection points and the detection parameters are determined, the automatic quality detection device is installed at each detection point, and the yarn appearance image data and the mechanical property data are collected in real time; The collected yarn appearance image data is subjected to image preprocessing to extract the yarn surface defect features, the defect recognition and classification are performed according to the preset defect types and thresholds, and the yarn appearance quality evaluation result is obtained. The yarn mechanical property data collected are subjected to feature extraction and statistical analysis, and the quality is evaluated according to preset mechanical property indexes and thresholds to obtain the yarn mechanical quality evaluation result; The yarn appearance quality evaluation result and the mechanical quality evaluation result are fused to construct a multi-source quality data set, and the Kalman filtering algorithm is used to correlate and fuse the multi-source quality data to obtain a comprehensive quality evaluation result; According to the yarn production process and the sequence of each process, the comprehensive quality evaluation result is arranged in chronological order to construct a yarn quality traceability chain, realizing the whole-process tracking and management of yarn quality; When the yarn winding problem is detected, the time node of the problem is traced back through the quality traceability chain, and the quality data and process parameters at the time node are used to analyze the root cause by using machine learning algorithm, and the key factors causing the yarn winding problem are located; According to the positioning result of the yarn winding problem, the process parameters of the corresponding process are optimized and adjusted, and the adjusted process parameters are verified and evaluated, and the optimization is iterated until the yarn winding problem is eliminated and the yarn quality is stable and controllable.

8. A yarn untwisting method according to claim 1, wherein The yarn production whole-process quality control system is established, and multi-source heterogeneous data of yarn raw materials, process, equipment and quality are integrated; principal component analysis and neural network algorithm are used to mine the complex correlation between yarn quality influencing factors, and a yarn quality optimization decision scheme is formed; According to the optimization scheme, the production parameters are automatically adjusted to continuously improve the yarn quality, improve the production efficiency and economic benefit, including: Obtaining multi-source heterogeneous data of raw materials, process, equipment and quality in the yarn production process, and converting them into structured data through data cleaning, conversion and integration; For the integrated structured data, principal component analysis algorithm is used to extract the main influencing factors of yarn quality and reduce the data dimension; According to the results of principal component analysis, a neural network model is constructed to deeply mine the complex nonlinear relationship between each influencing factor and yarn quality; Through the trained neural network model, the yarn quality under different production parameter combinations is predicted, and the optimal parameter combination is selected as the quality optimization scheme; The quality optimization scheme is converted into control instructions executable by production equipment, and the production parameters are adjusted in real time through an automatic control system; The optimized production data are continuously collected, and the principal component analysis and neural network model are updated regularly to form a closed-loop control for continuous optimization of yarn quality; A yarn quality control visualization platform is established to monitor the key quality indexes of each production link in real time and provide data support for management decision-making.

Images