Large-model-assisted dynamic configuration and quick response method for knitting production line

Abstract

The invention provides a large-model-assisted dynamic configuration and quick response method for a knitting production line, and the method comprises the steps: obtaining the historical data of the knitting production line according to the dye molecular weight, the dye polar group number and the dye molecular space structure as dye property parameters, and constructing a dye property data warehouse; according to historical data of a knitting production line, dye molecular structure parameters, yarn surface morphology, material characteristics and friction coefficient test data are selected, and a correlation prediction model of dye properties and yarn friction coefficients is constructed; according to dye properties and friction coefficient distribution characteristics, performing clustering grouping on dye formula compositions, determining dye combinations having similar influence on yarn friction coefficients, and forming a dye formula and friction coefficient mapping library; knitting mechanical structure parameters, a knitting needle motion trail equation, a fabric organization structure matrix and fabric density parameters are obtained and combined with a dye formula and friction coefficient mapping library, and a weaving process digital twinning model is constructed.

Description

Technical Field

[0001] The invention relates to the field of information technology, and in particular to a large model-assisted dynamic configuration and rapid response method for a knitting production line. Background Art

[0002] In knitting production lines, dyes of different colors have subtle but critical effects on the friction coefficient of yarns. Slight changes in the friction coefficient of yarns can lead to differences in yarn tension and speed during the weaving process, which in turn affects the uniformity, density and texture of the fabric. Traditional production methods make it difficult to accurately predict and control the changes in the friction coefficient of yarns caused by different dyes, which often leads to large differences between the finished products and design expectations, affecting product quality and production efficiency. The dye formula of the dye has a complex effect on the friction behavior between the yarn and the loom components. The core of this problem lies in how to accurately capture and characterize the influence of dye factors on the friction coefficient of yarns, and effectively integrate it into the weaving process model, which are key issues that need to be solved urgently. At the same time, how to balance the complexity and computational efficiency of the model while ensuring the accuracy of the model is also an important challenge. Summary of the invention

[0003] The present invention provides a large model-assisted knitting production line dynamic configuration and rapid response method, which mainly includes:

[0004] Based on the dye molecular weight, the number of dye polar groups, and the spatial structure of dye molecules as dye property parameters, the historical data of knitting production lines are obtained and a dye property data warehouse is constructed;

[0005] Based on the historical data of knitting production lines, dye molecular structure parameters, yarn surface morphology, material properties and friction coefficient test data were selected to construct a correlation prediction model between dye properties and yarn friction coefficient;

[0006] The correlation data between dye properties and yarn friction coefficients were mined, and the correlation between dye molecular structure parameters and yarn surface morphology, material properties and friction coefficient numerical ranges was analyzed to obtain the distribution characteristics of yarn dye properties and friction coefficients under different dye property combinations;

[0007] According to the dye properties and friction coefficient distribution characteristics, the dye formula composition is clustered and grouped, and the dye combination with similar influence on the yarn friction coefficient is determined to form a dye formula and friction coefficient mapping library;

[0008] Acquire knitting machine structure parameters, needle motion trajectory equations, fabric structure matrix and fabric density parameters, and combine them with dye formula and friction coefficient mapping library to build a digital twin model of the weaving process;

[0009] The dye properties, yarn characteristics and friction coefficient boundary conditions are set in the digital twin model, and the finite element simulation method is used to simulate the yarn movement and fabric formation process during weaving, and the influence of the dye properties and the spinning friction coefficient during weaving on the weaving process is analyzed;

[0010] During the weaving process, the data of the friction coefficient of the dye properties and yarn are collected in real time, and the simulation results of the digital twin model are compared to dynamically adjust the needle motion trajectory equation and weaving parameters to achieve real-time optimization control of the weaving process;

[0011] When a new dye formula is used in a knitting production line, the yarn friction coefficient is predicted based on the dye formula and the friction coefficient mapping library, and the weaving process is simulated and optimized through the digital twin model to provide weaving parameter setting ranges and weaving process optimization suggestions;

[0012] Continuously collect data on dye properties, yarn friction coefficient and weaving process parameters, iteratively update and adaptively adjust the digital twin model and weaving process optimization system, improve the stability of weaving quality and the process adaptability of new dye formulas, and match dye formulas with weaving parameters.

[0013] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0014] The present invention discloses a large-model-assisted dynamic configuration and rapid response method for a knitting production line. The method realizes precise matching of dye properties, yarn characteristics and weaving parameters by constructing a dye property data warehouse and a dye formula and friction coefficient mapping library, combined with a digital twin model of the weaving process. The present invention first analyzes the correlation between the dye molecular structure and the yarn friction coefficient, then uses digital twin technology to simulate the weaving process, and dynamically adjusts the weaving parameters through machine learning. When using a new dye formula, the present invention can predict the yarn friction coefficient and give weaving process optimization suggestions. Through continuous data collection and model iteration, the present invention continuously improves the stability of weaving quality and the process adaptability of new dye formulas, effectively solves the problem of the impact of dye property changes on knitting production, and realizes the intelligent and refined control of knitting production. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 The present invention is a flowchart of a large-model-assisted knitting production line dynamic configuration and rapid response method.

[0016] Figure 2 It is a schematic diagram of a large-model-assisted knitting production line dynamic configuration and rapid response method of the present invention.

[0017] Figure 3 It is another schematic diagram of a large-model-assisted knitting production line dynamic configuration and rapid response method of the present invention. DETAILED DESCRIPTION

[0018] 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.

[0019] like Figure 1-3 In this embodiment, a large model-assisted knitting production line dynamic configuration and rapid response method may specifically include:

[0020] Step S101, based on the dye molecular weight, the number of dye polar groups, and the dye molecular spatial structure as dye property parameters, the knitting production line historical data is acquired to build a dye property data warehouse.

[0021] The method comprises the following steps: obtaining the molecular weight information of the dye and the number of polar groups, wherein the molecular weight information of the dye is obtained from the PubChem database, and the number of polar groups is calculated by the SMILES string analysis method; constructing a three-dimensional spatial structure model of the dye molecule according to the molecular weight information of the dye and the number of polar groups by using the PyMOL molecular modeling tool; extracting the production batch information in the past five years from the production line history database, and performing outlier detection and processing on the production batch information by using the Python Pandas library to obtain a standardized data set; constructing a data warehouse by using Apache Hive, associating and mapping the dye property parameters with the standardized data set to fact tables and dimension tables, and constructing a multidimensional data cube by using Microsoft Analysis Services; analyzing the relationship between the dye properties and the production parameters by using the correlation analysis method, and constructing a dye selection prediction model by using the CART decision tree algorithm; and if it is necessary to generate a dye selection recommendation list, extracting the association rules between the dye properties and the production parameters from the data warehouse by using the Apriori association rule mining algorithm.

[0022] Specifically, the molecular weight information of dyes was obtained from PubChem database, the number of polar groups of dyes was calculated and recorded by SMILES string analysis method, and the three-dimensional spatial structure model of dye molecules was constructed by PyMOL molecular modeling tool and the coordinate data was saved. The production batch information in the past five years was extracted from the historical database of knitting production line, the data was classified and sorted according to the production process parameters, and the outlier detection and processing were performed using Python's Pandas library. After removing duplicate items, a standardized data set was formed. Apache Hive was used to build a data warehouse, and the dye property parameters were associated with the historical data of the production line and mapped to the fact table and dimension table. Microsoft Analysis Services was used to build a multidimensional data cube, and index optimization was used to improve data retrieval efficiency. The relationship between dye properties and production parameters was deeply analyzed using the correlation analysis method, and the dye selection prediction model was constructed using the CART decision tree algorithm. The dye property data warehouse query interface was established for the production system to call. The Apriori association rule mining algorithm was used to extract the association rules between dye properties and production parameters from the data warehouse to generate a dye selection recommendation list. In the data collection stage, the dye molecular structure information was queried through the PubChem API interface to obtain molecular weight data. Taking azo dye as an example, its molecular weight is 327.33 g / mol. The SMILES string "N=Nc1ccc(cc1)S(=O)(=O)O" was parsed using the RDKit library to identify polar groups containing nitrogen, oxygen, and sulfur, and the number of polar groups was calculated to be 5. The molecular structure file was imported using PyMOL software to generate a three-dimensional model and extract atomic coordinate data, such as [(1.23,-0.45,0.78),(2.56,0.89,-1.34)]. Production records from January 2019 to January 2024 were extracted from the knitting production line database, including parameters such as dye dosage, temperature, and time. The interquartilerange method of the Pandas library was used to detect outliers, such as marking data with dye dosage exceeding 3 times the standard deviation as abnormal. Duplicate records were deleted using the drop_duplicates function, and a total of 50,000 standardized data sets were finally obtained. Use Apache Hive to create a data warehouse and design a star schema. Use dye batches as fact tables to connect dimension tables such as dye attributes, production parameters, and time. Use Microsoft Analysis Services to build an OLAP multidimensional data cube for sales, production, and quality, and optimize query performance through bitmap indexes. Use the Pearson correlation coefficient to analyze the relationship between dye polarity and dyeing fastness, and the calculated correlation coefficient is 0.78, indicating a strong correlation.Using the CART decision tree algorithm, with dye molecular weight, polar group number, and spatial structure characteristics as input variables and dyeing effect as the target variable, a prediction model was constructed with an accuracy of 85%. Finally, the Apriori algorithm was used to mine association rules, with the minimum support set to 0.1 and the minimum confidence set to 0.7. Ten valid rules such as "high number of polar groups → high dyeing fastness" were found from 50,000 records, generating a dye selection recommendation list.

[0023] Step S102, based on the historical data of the knitting production line, select the dye molecular structure parameters, yarn surface morphology, material properties and friction coefficient test data to build a correlation prediction model between the dye properties and the yarn friction coefficient.

[0024] The molecular structural parameters of dyes are obtained, including molecular weight, number of polar groups and spatial configuration; microscopic photos of yarn surface are analyzed to obtain fiber arrangement and surface roughness, and yarn material property data and friction coefficient are collected. The original data is preprocessed, and if outliers are detected, they are eliminated. The spline interpolation algorithm is used to complete the missing data, and the data from different sources are formatted and converted to obtain processed data. A multidimensional data model is constructed, in which dye batches and yarn batches are used as fact tables, and dye molecular structural parameters, yarn surface morphology, material properties and friction coefficient data are used as dimension tables; a time series analysis is performed on historical data to obtain the dynamic changes in the relationship between dye properties and yarn friction coefficient. Dimensionality reduction is performed, eigenvalues ​​are calculated, and the principal components whose cumulative contribution rate reaches the preset threshold are selected to obtain dimension-reduced data. A correlation prediction model between dye properties and yarn friction coefficient is constructed, and the number of decision trees is set; the Gini index is used for feature selection, and the accuracy of the model is evaluated through cross-validation to obtain optimization suggestions for dye selection and yarn friction coefficient.

[0025] Specifically, the molecular structure parameters of dyes, including molecular weight, number of polar groups and spatial configuration, were obtained from the molecular structure database. ImageJ software was used to analyze the microscopic photos of the yarn surface, and surface morphological characteristic values ​​such as fiber arrangement and surface roughness were extracted. The material property data of the yarns were collected through material testing instruments, and the friction coefficient of the yarns was measured using a friction coefficient tester. The collected raw data were preprocessed, the outliers were marked and removed, the spline interpolation algorithm was used to fill in the missing data, the format of data from different sources was unified and the units were converted, and finally the processed data were stored in the MySQL database. A multidimensional data model was constructed using a star schema, with dye batches and yarn batches as fact tables, and dye molecular structure parameters, yarn surface morphology, material properties and friction coefficient data as dimension tables, mapped to a multidimensional data cube, and a time series analysis of historical data was performed to capture the dynamic changes in the relationship between dye properties and yarn friction coefficient. The principal component analysis method was used to reduce the dimensionality of multidimensional data, calculate the eigenvalues ​​and select the principal components with a cumulative contribution rate of 85%. The association prediction model between dye properties and yarn friction coefficient was constructed by random forest algorithm, 500 decision trees were set, and the Gini index was used for feature selection. The accuracy of the model was evaluated by ten-fold cross validation, and dye selection and yarn friction coefficient optimization suggestions were output. The molecular structure parameters of the dye were obtained from the ChemSpider database, such as the molecular weight of azo dyes was 327.33g / mol, the number of polar groups was 5, and the spatial configuration was a planar structure. The microscopic photos of the yarn surface were analyzed using ImageJ software, the threshold was set to 128, the fiber arrangement was extracted to be 0.85, and the surface roughness was 15μm. The material testing instrument collected the yarn material property data, and the fiber length was 38mm and the fiber fineness was 1.5 denier. The friction coefficient tester measured the yarn friction coefficient to be 0.32. The raw data were preprocessed, and the outliers were marked and removed using the 3-times standard deviation method. The missing data were supplemented by the cubic spline interpolation algorithm. All length units were unified into millimeters and mass units were unified into grams. The processed data were stored in the MySQL database, and four tables were created: dye batch, yarn batch, dye attribute, and yarn attribute. The multidimensional data model was constructed using a star schema, with dye batch and yarn batch as fact tables, and four dimensional tables: dye molecular structure, yarn surface morphology, material properties, and friction coefficient were connected. Time series analysis was performed on five years of historical data, and the trend of the relationship between dye properties and yarn friction coefficient was predicted using the exponential smoothing method. In the principal component analysis, the eigenvalues ​​were calculated and the first three principal components with a cumulative contribution rate of 85% were selected. The random forest algorithm set 500 decision trees, and the Gini index was used for feature selection, with 3 features randomly selected at each node. The model was evaluated using ten-fold cross validation, and the average accuracy was 87.5%.The final model outputs dye selection recommendations. For given yarn parameters, dyes with a molecular weight in the range of 300-350 g / mol and 4-6 polar groups are recommended, and the friction coefficient is predicted to be reduced to 0.28.

[0026] Step S103, mining the correlation data between dye properties and yarn friction coefficient, analyzing the correlation between dye molecular structure parameters and yarn surface morphology, material properties and friction coefficient numerical range, and obtaining the dye properties and friction coefficient distribution characteristics of the yarn under different dye property combinations.

[0027] The dye molecular structure parameters, yarn surface morphology, material properties and friction coefficient data in the data warehouse are obtained; according to the dye molecular structure parameters, yarn surface morphology, material properties and friction coefficient data, the Apriori algorithm is used to mine association rules to obtain a set of strong association rules between the dye molecular structure parameters and the yarn surface morphology, material properties and friction coefficient; for the set of strong association rules, the K-means clustering algorithm is used to classify the dye property combinations, and the statistical distribution parameters of the yarn friction coefficient under each category are determined; if the statistical distribution parameters meet the preset conditions, the multivariate linear regression method is used to establish a mathematical model between the dye properties and the yarn friction coefficient; through the mathematical model, the friction coefficient distribution characteristics of the yarn under different dye property combinations are obtained; according to the friction coefficient distribution characteristics, the box plot and kernel density estimation method are used to generate visualization results; it is judged whether the visualization results meet the preset threshold, and if so, the dye property and friction coefficient distribution characteristics report is output.

[0028] Specifically, the molecular structure parameters of dyes, yarn surface morphology, material properties and friction coefficient data were extracted from the data warehouse, and the outliers were processed using the 3-times standard deviation method. The data of different dimensions were converted into a unified scale through the Z-score standardization method. The parameters such as the molecular weight of dyes and the number of polar groups were discretized with equal width and divided into 10 intervals. Finally, the processed data set was divided into a training set and a test set in a ratio of 7:3. The Apriori algorithm was used to mine association rules for the data set, with the minimum support set of 0.05 and the minimum confidence set of 0.6. The association rules between the molecular structure parameters of dyes and the surface morphology of yarns, material properties and friction coefficients were mined to obtain a set of strong association rules. According to the association rules mined, the K-means clustering algorithm was used to classify the dye property combinations, and the optimal number of clusters was determined by the silhouette coefficient method. The statistical distribution parameters such as the mean, standard deviation, and quartile of the yarn friction coefficient under each category were calculated. The mathematical model between dye properties and yarn friction coefficient was established using the multivariate linear regression method. The model was verified and parameters were optimized using 10-fold cross validation. The mapping relationship between the dye property combination and the distribution characteristics of the yarn friction coefficient was obtained. The friction coefficient distribution characteristics of the yarn under different dye property combinations were visualized through box plots and kernel density estimation, and a report on the dye property and friction coefficient distribution characteristics was output.

[0029] 100,000 records were extracted from the data warehouse, including data such as dye molecular weight (range 200-800 g / mol), number of polar groups (0-10), yarn surface roughness (1-20 μm) and friction coefficient (0.1-0.5). The 3-times standard deviation method was used to identify outliers and 374 records were removed. After Z-score standardization, the data mean was 0 and the standard deviation was 1. Parameters such as dye molecular weight were divided into 10 equal width intervals, such as molecular weight intervals [200-260), [260-320), etc. The training set contained 69,639 records and the test set contained 29,987 records. Apriori algorithm was used to mine association rules, with the minimum support of 0.05 and the minimum confidence of 0.6. 152 strong association rules were obtained, such as {molecular weight: [320-380), polar group: [4-5)}=>{friction coefficient: [0.25-0.30)}, with a support of 0.068 and a confidence of 0.72. K-means clustering divided the dye property combinations into 5 categories, and the optimal number of clusters was determined by the silhouette coefficient method (silhouette coefficient 0.68). The statistical parameters of the friction coefficient of each category were calculated, such as the mean of category 1 was 0.28, the standard deviation was 0.03, Q1 was 0.26, and Q3 was 0.30. Multiple linear regression modeling was performed, with the independent variables being the molecular weight of the dye, the number of polar groups, etc., and the dependent variable being the friction coefficient. The average R was obtained by 10-fold cross validation. 2The model shows that for every 100 g / mol increase in molecular weight, the friction coefficient decreases by 0.015 on average; for every increase in polar groups, the friction coefficient increases by 0.008 on average. The box plot shows the distribution of friction coefficients in five categories, and the kernel density estimation plot shows that the friction coefficient distributions of category 1 and category 3 have a high degree of overlap, which may require further subdivision. The final report contains 152 association rules, five dye property combination categories and their corresponding friction coefficient distribution characteristics.

[0030] The correlation between the molecular structure and chromatographic characteristics of the dye and the surface roughness and linear density of the yarn was analyzed to obtain the surface morphological characteristics of the yarn under different dye properties.

[0031] The molecular structural parameters of the dye are obtained, the absorption spectrum and reflectance spectrum of the dye are measured, and standardized chromatographic characteristic data are obtained; the surface roughness and linear density of the yarn are measured to form a yarn surface morphological characteristic data set; the dimensionality reduction processing is performed on the molecular structural parameters of the dye and the chromatographic characteristic data set to extract the main characteristic variables; the correlation coefficient between the main characteristic variables and the yarn surface morphological characteristic data set is calculated, and the significantly correlated feature combination is screened out according to the correlation coefficient; a prediction model between the dye properties and the yarn surface morphological characteristics is established using the significantly correlated feature combination; the prediction model is used to output the surface roughness and linear density prediction values ​​of the yarn under different dye property combinations; the predicted values ​​are clustered and analyzed to obtain multiple categories; the center point and standard deviation of each category are calculated, the characteristic description of each category is given, a scatter plot is used to visualize the relationship between the dye properties and the yarn surface morphological characteristics, and a heat map is used to display the distribution of the yarn surface morphological characteristics under different dye property combinations.

[0032] Specifically, the molecular structure parameters of the dye are obtained from the dye database, the absorption spectrum of the dye is measured by an ultraviolet-visible spectrophotometer, the reflectance spectrum is measured by a spectrophotometer, and the raw data is denoised and baseline corrected by data preprocessing software to obtain standardized chromatographic characteristic data. The surface roughness of the yarn is measured by a laser roughness meter, and the linear density value is directly measured by an electronic yarn density meter. The measurement results are transmitted to a computer through a data acquisition card to form a yarn surface morphological characteristic data set. The principal component analysis method is used to reduce the dimension of the dye molecular structure and chromatographic characteristic data, and the cumulative explained variance ratio method is used to determine the number of retained principal components, extract the main characteristic variables, and calculate the correlation between the dye characteristics and the yarn surface roughness and linear density values ​​by the Pearson correlation coefficient, and screen out significantly correlated feature combinations. According to the correlation analysis results, the support vector machine regression algorithm is used to establish a prediction model between dye properties and yarn surface morphological characteristics. The radial basis function kernel is used to optimize the model parameters by the grid search method, and the model performance is evaluated by cross-validation. The surface roughness and linear density prediction values ​​of the yarn under different dye property combinations are output. The prediction results were divided into multiple categories using the K-means clustering algorithm, and the center point and standard deviation of each category were calculated, and the characteristic description of each category was given. The relationship between the dye properties and the yarn surface morphological characteristics was visualized using a scatter plot, and the distribution of the yarn surface morphological characteristics under different dye property combinations was displayed using a heat map. The molecular structure parameters of 100 dyes were extracted from the dye database, with a molecular weight range of 200-800g / mol, a number of polar groups of 0-10, and spatial configurations including planar and stereoscopic structures. The absorption spectrum was measured at 2nm intervals in the range of 200-800nm ​​using a UV-visible spectrophotometer, and the reflectance spectrum was measured at 10nm intervals in the range of 400-700nm using a spectrophotometer. The data preprocessing software used wavelet transform for denoising and polynomial fitting for baseline correction. The laser roughness meter measured the yarn surface roughness with a resolution of 0.1μm and a range of 1-20μm. The electronic yarn density meter measured the linear density value with an accuracy of 0.1tex and a range of 10-50tex. The principal component analysis retains principal components with a cumulative explained variance of 95%, usually 5-8. The threshold of the Pearson correlation coefficient is set to 0.6, and 30 significantly correlated feature combinations are screened out. The support vector machine regression model uses the RBF kernel, and the parameters C (penalty coefficient) and γ (kernel function coefficient) are optimized through grid search. The range of C is [0.1, 1, 10, 100], and the range of γ is [0.01, 0.1, 1, 10]. The model performance is evaluated by 5-fold cross validation, and the final model R 2Reached 0.85. K-means clustering divides the prediction results into 5 categories, and calculates the center point and standard deviation of each category, such as category 1: surface roughness 12.5±1.2μm, linear density 30.5±2.3tex. The scatter plot shows the relationship between the molecular weight of the dye and the surface roughness of the yarn, and the heat map shows the influence of the number of polar groups and spatial configuration on the linear density. The color depth indicates the size of the linear density value.

[0033] Step S104, clustering and grouping the dye formula compositions according to the dye properties and the friction coefficient distribution characteristics, determining dye combinations with similar effects on the yarn friction coefficient, and forming a dye formula and friction coefficient mapping library.

[0034] Dye property data and yarn friction coefficient data were obtained, and the Z-score standardization method was used to convert data of different dimensions into a unified scale. Principal component analysis was performed on the standardized data, and the principal component with a cumulative contribution rate of 95% was determined as the input feature for cluster analysis. The DBSCAN algorithm was used to preprocess the input features to obtain a data set with outliers removed. The dye formula composition was grouped using the K-means clustering algorithm according to the data set, and the optimal number of clusters was determined by the silhouette coefficient method. If the optimal number of clusters was determined, the dye property parameters and the corresponding friction coefficient mean of each cluster center were calculated. Based on the clustering results, a mapping relationship between dye formula and friction coefficient was constructed. The CART decision tree algorithm was used to establish a prediction model from dye properties to friction coefficient, and the prediction model parameters were optimized through grid search. A dye formula and friction coefficient mapping library was generated based on the prediction model, and the mapping library contained dye combination features and corresponding friction coefficient prediction values. The sliding window method was used to update the mapping relationship between dye formula and friction coefficient in the mapping library.

[0035] Specifically, the dye property data and the corresponding yarn friction coefficient data were extracted from the data warehouse, including the molecular weight, number of polar groups, spatial configuration and other parameters of single dyes and mixed dyes. The outliers were processed by Python's Pandas library, and the data of different dimensions were converted into a unified scale using the Z-score standardization method to form a standardized dye property and friction coefficient data set. The principal component analysis method was used to reduce the dimension of the dye property data, extract the main characteristic variables, calculate the contribution rate of each principal component, and select the principal component with a cumulative contribution rate of 95% as the input feature of the cluster analysis. The DBSCAN algorithm was used to preprocess the data to remove outliers, and then the K-means clustering algorithm was used to group the dye formula composition. The optimal number of clusters was determined by the silhouette coefficient method, and the dye property parameters and the corresponding friction coefficient mean of each cluster center were calculated to obtain dye combinations with similar effects on the yarn friction coefficient. Based on the clustering results, the mapping relationship between dye formula and friction coefficient was constructed. The CART decision tree algorithm was used to establish a prediction model from dye properties to friction coefficient. The parameters were optimized through grid search, including the depth of the tree and the minimum number of samples of leaf nodes. The model performance was evaluated using 10-fold cross validation. A mapping library of dye formula and friction coefficient was generated, which contained dye combination characteristics and corresponding friction coefficient prediction values. The mapping relationship was updated monthly using the sliding window method with a window size of 6 months. A heat map was used to visualize the relationship between dye combination and friction coefficient. The horizontal axis represented different dye combinations, the vertical axis represented the friction coefficient range, and the color depth represented the frequency of occurrence. 10,000 dye property and friction coefficient data were extracted from the data warehouse, including single dyes and two-component mixed dyes, with a molecular weight range of 200-800 g / mol, the number of polar groups of 0-10, and the spatial configuration was divided into planar and three-dimensional. The interquartilerange method of the Pandas library was used to detect outliers and 374 records were eliminated. After Z-score standardization, the data mean was 0 and the standard deviation was 1. The results of principal component analysis showed that the cumulative contribution rate of the first five principal components reached 96.3%, which was selected as clustering input. The parameters of the DBSCAN algorithm were set to ε = 0.5, MinPts = 5, and 86 outliers were identified and removed. K-means clustering tried 2-10 clusters, and the optimal number of clusters was determined to be 6 by the silhouette coefficient, and 6 types of dye combinations with similar effects on the friction coefficient were obtained. The grid search range of the CART decision tree parameters was tree depth [3, 6, 9], minimum number of leaf node samples [5, 10, 20], and 10-fold cross validation average R 2The generated mapping library contains 6,000 dye formulas and friction coefficient correspondences. The sliding window is updated monthly, processing about 3,000 new data in the last 6 months. The heat map visualization results show that the horizontal axis is 6 types of dye combinations, the vertical axis friction coefficient range is 0.1-0.5, and the color from light blue to dark red indicates the frequency from low to high. Among them, the third type of dye combination, molecular weight 400-600g / mol, 6-8 polar groups, stereo configuration, is highly correlated with low friction coefficient, 0.15-0.25, and appears dark red.

[0036] Step S105, obtaining knitting machine structure parameters, knitting needle motion trajectory equations, fabric structure matrix and fabric density parameters, combining them with dye formula and friction coefficient mapping library, to build a digital twin model of the weaving process.

[0037] The mechanical structure parameters in the knitting machinery database are obtained, and the mechanical structure parameters include the needle cylinder diameter, the number of needles and the coil length; the high-precision position sensor is used to collect the knitting needle movement position information, and the knitting needle movement trajectory equation is obtained by fitting the position information through the cubic spline interpolation method. According to the fabric design drawings, the convolutional neural network is used for image segmentation and feature extraction to obtain the fabric structure matrix; the fabric density parameters are obtained through the density measuring instrument, and the structure matrix and density parameters are stored in the database. The random forest regression method is used to establish the association model between the dye formula, the friction coefficient and the weaving parameters; the data in the dye formula and friction coefficient mapping library are matched with the weaving parameters to obtain a comprehensive data set. The long short-term memory network is used to construct a digital twin model of the weaving process, and the digital twin model takes the mechanical parameters, trajectory equation, structure matrix, density parameters and dye formula as input and the friction coefficient as output; the parameters of the digital twin model are optimized by the back propagation algorithm. The operation process of the digital twin model is visualized by three-dimensional rendering technology, and the knitting needle movement and fabric formation process are displayed in real time; the digital twin model is optimized by the particle swarm optimization algorithm to improve the prediction accuracy.

[0038] Specifically, mechanical structure parameters, including needle cylinder diameter, number of needles, coil length and other data, are obtained from the knitting machinery database. High-precision position sensors are used to collect knitting needle motion position information. The knitting needle motion trajectory equation is fitted by cubic spline interpolation method, and the mechanical parameters and trajectory equation are stored in the database. According to the fabric design drawings, convolutional neural networks are used for image segmentation and feature extraction to construct a fabric structure matrix. The fabric density parameters are obtained by density measuring instruments, and the structure matrix and density parameters are stored in the database. The random forest regression method is used to establish an association model between dye formula, friction coefficient and weaving parameters. The data in the dye formula and friction coefficient mapping library are matched with the weaving parameters to obtain a comprehensive data set. The long short-term memory network is used to construct a digital twin model of the weaving process. The mechanical parameters, trajectory equations, structure matrix, density parameters and dye formula are used as inputs, and the friction coefficient is used as output. The model parameters are optimized by the back propagation algorithm to realize the digital representation of the weaving process. The model is verified using actual weaving data, and the model is optimized by the particle swarm optimization algorithm to improve the prediction accuracy. The operation process of the digital twin model is visualized using 3D rendering technology, showing the movement of knitting needles and the process of fabric formation in real time.

[0039] The mechanical structure parameters were extracted from the knitting machinery database. The needle cylinder diameter was 18 inches, the number of needles was 2400, and the coil length was 2.5 mm. The high-precision position sensor collected the knitting needle movement position information at a frequency of 1kHz and obtained 10,000 data points. The cubic spline interpolation method was used to fit the knitting needle movement trajectory equation, and the fitting error was less than 0.01 mm. The convolutional neural network used the ResNet-50 architecture to segment and extract features from the 1024x1024 pixel fabric design image, with an accuracy of 98%. The density meter measured the fabric density to be 320 g / m 2 . The random forest regression model uses 100 decision trees, and feature importance analysis shows that the molecular weight of the dye has the greatest impact on the friction coefficient, with a contribution rate of 35%. The long short-term memory network contains 3 hidden layers, 128 neurons in each layer, and the input sequence length is 50. The parameters are adjusted by the Adam optimizer with a learning rate of 0.001. The model was verified on 10,000 actual weaving data with a mean absolute error of 0.015. The particle swarm optimization algorithm uses 50 particles and iterates 100 times, reducing the model prediction error by 12%. The 3D rendering technology is implemented using OpenGL to visualize the digital twin model at a speed of 60 frames per second, showing the real-time movement of 2,400 knitting needles and the fabric formation process, with a texture accuracy of 0.1mm. The entire digital twin model runs on a workstation equipped with an RTX 3080 graphics card, and the simulation time for a single weaving process is 0.5 seconds.

[0040] Step S106, setting the dye properties, yarn characteristics and friction coefficient boundary conditions in the digital twin model, using the finite element simulation method to simulate the yarn movement and fabric formation process during weaving, and analyzing the influence of the dye properties and the spinning friction coefficient during weaving on the weaving process.

[0041] Dye properties, yarn characteristics and friction coefficient data are obtained, and the boundary condition range is set, wherein the boundary condition range includes the upper limit value of dye molecular weight, the lower limit value of dye molecular weight, the upper limit value of polar group number, the lower limit value of polar group number, the upper limit value of yarn linear density, the lower limit value of yarn linear density, the upper limit value of yarn strength, the lower limit value of yarn strength, the upper limit value of yarn elongation, the lower limit value of yarn elongation, the upper limit value of friction coefficient and the lower limit value of friction coefficient. The yarn geometric model and the fabric geometric model are constructed by ANSYS finite element analysis software. If the yarn geometric model is constructed, the yarn geometric model is discretized into beam units, wherein each yarn is divided into a preset number of units. The yarn material properties and contact conditions are determined according to the yarn geometric model, and the knitting needle motion trajectory and weaving tension are set as external loads. The weaving process is simulated by an explicit dynamics solver, which determines the time step based on the Courant-Friedrichs-Lewy condition. The experimental scheme was set up by orthogonal experimental design method, and the experimental scheme included the number of factors and the number of levels. The simulation experiment was carried out according to the experimental scheme, and the significance test of the simulation results was carried out by variance analysis. ParaView scientific visualization software was used to display the yarn motion trajectory and stress distribution, and the weight model of influencing factors was established by grey correlation analysis. The weight model of influencing factors was used to quantitatively describe the influence of each factor on the weaving process.

[0042] Specifically, the dye properties, yarn characteristics and friction coefficient data are extracted from the database, and the boundary condition range is set, including dye molecular weight, number of polar groups, yarn linear density, strength, elongation and upper and lower limits of friction coefficient. These parameters are imported into the digital twin model as initial conditions. The geometric model of yarn and fabric is constructed using ANSYS finite element analysis software, the yarn is discretized into beam units, each yarn is divided into 500 units, the yarn material properties and contact conditions are defined, the knitting needle motion trajectory and weaving tension are set as external loads, and a complete finite element simulation model is established. The explicit dynamics solver is used to simulate the weaving process. The time step is determined to be 1e-6 seconds based on the Courant-Friedrichs-Lewy condition. The movement and deformation of the yarn under the action of the knitting needle are simulated, the position, stress and strain of the yarn at each time step are calculated, and the changes in key parameters during the fabric formation process are recorded. Through the orthogonal experimental design method, a test scheme with 3 factors and 5 levels was set up, 125 simulation experiments were carried out, and the significance test of the simulation results was carried out by variance analysis. The quadratic polynomial response surface equation was established to obtain the influence of dye properties and friction coefficient on the weaving process. ParaView scientific visualization software was used to display the yarn motion trajectory and stress distribution. The weight model of the influencing factors was established by the grey correlation analysis method to quantitatively describe the influence of each factor on the weaving process. The simulation accuracy was verified by comparing with the actual weaving experimental results, and the error was controlled within 5%. 100 kinds of dye property data were extracted from the database, with a molecular weight range of 200-800g / mol and a number of polar groups of 1-10; yarn properties include linear density of 10-50tex, strength of 3-8cN / dtex, elongation of 5-20%; friction coefficient range of 0.1-0.5. A circular knitting machine model with a diameter of 18 inches and a needle number of 2400 was constructed in ANSYS. The yarn used BEAM188 beam elements, each yarn was divided into 500 elements, and the element length was 0.5 mm. The material properties were set to linear elasticity, Young's modulus 1000 MPa, and Poisson's ratio 0.3. The knitting needle motion trajectory was set to a sine curve with a frequency of 600 rpm and an amplitude of 5 mm. The explicit dynamics solver time step was 1e-6 seconds, the total simulation time was 0.1 seconds, and 10,000 frames of data were output. The orthogonal experimental design had 3 factors, molecular weight, number of polar groups, and friction coefficient, 5 levels, and a total of 125 groups of experiments. Variance analysis showed that molecular weight and friction coefficient had significant effects on the weaving process (P<0.01). The fit of the quadratic polynomial response surface equation R 2 =0.95. ParaView visualization shows that the maximum yarn stress occurs at the contact point with the knitting needle, with a peak stress of 80MPa. Grey correlation analysis yields the influence weights, with a friction coefficient of 0.42, a molecular weight of 0.35, and a number of polar groups of 0.23. Compared with the actual weaving experiment, the simulation predicted fabric density error of 4.2% and the coil length error of 3.8%, both within the acceptable range.

[0043] The molecular size, polarity, adsorption capacity parameters of the dye and the friction coefficient of the yarn were set as boundary conditions. Finite element simulation was used to simulate the dynamic changes of yarn tension, speed and needle motion trajectory during the weaving process, and the influence of dye properties and friction coefficient on the stability of the weaving process was analyzed.

[0044] The molecular size, polarity and adsorption capacity data of dyes are obtained from a pre-established dye database; the friction coefficient of yarns is measured by the inclined plane method, and the measurement is carried out under preset temperature and humidity conditions. According to the obtained dye data and friction coefficient, a numerical model of the weaving process is constructed; wherein the yarn is discretized into predefined beam units, and the knitting needle motion trajectory equation is set. The weaving process is simulated using an explicit dynamics solver; the tension, velocity and displacement of the yarn at each time step are calculated, and the real-time coordinates of the knitting needle motion trajectory are recorded. Through the orthogonal experimental design method, a multi-factor and multi-level experimental scheme is set; multiple simulation experiments are carried out to calculate the fluctuation coefficient of yarn tension, velocity and knitting needle motion trajectory. If the fluctuation coefficient is less than the preset threshold, the weaving process is determined to be stable; otherwise, a sensitivity analysis is performed to determine the degree of influence of each parameter on stability.

[0045] Specifically, the molecular size, polarity and adsorption capacity data of dyes were obtained from the dye database, and the friction coefficient of yarns was measured by the tilted plane method. The test conditions were temperature 20°C, relative humidity 65%, and tilt angle 0-90°. The outliers were processed using the 3-times standard deviation method, and the processed data were input into the ANSYS software as the boundary conditions of the finite element simulation. The numerical model of the weaving process was constructed using ANSYS, the yarn was discretized into BEAM188 beam elements, the knitting needle motion trajectory equation was set as a sine function, the friction contact conditions between the yarn and the knitting needle were defined, and the Maxwell-Wiechert viscoelastic yarn constitutive model considering the dye adsorption capacity was established. The weaving process was simulated using the central difference method in the explicit dynamics solver. The time step was set to 1e-6 seconds based on the Courant-Friedrichs-Lewy condition, and the total simulation time was 1 second. The tension, velocity and displacement of the yarn were calculated at each time step, and the real-time coordinates of the knitting needle motion trajectory were recorded. Through the orthogonal experimental design method, a 3-factor 5-level experimental scheme was set up, 125 simulation experiments were carried out, the simulation data were processed by the autoregressive moving average model, the fluctuation coefficients of yarn tension, speed and needle motion trajectory were calculated as quantitative indicators of weaving process stability, and the relationship model between dye properties and friction coefficient and weaving process stability was established by multivariate linear regression analysis. Sensitivity analysis was carried out to determine the influence of each parameter on stability, and the actual weaving experimental data were compared and verified, and the error was controlled within 5%. The molecular size (range 1-5nm), polarity (dipole moment 0.5-4D) and adsorption capacity (Gibbs free energy -10 to -50kJ / mol) data of 100 dyes were extracted from the dye database. The friction coefficient of the yarn was measured by the inclined plane method, and the inclination angle increased from 0° to 90° by 1° per second, and the angle at which the yarn started to slide was recorded. The average value was taken after 10 times of repetition, and the friction coefficient range was 0.1-0.5. A circular knitting machine model with a diameter of 18 inches and a needle number of 2400 was constructed in ANSYS. The yarn used BEAM188 units, and each yarn was divided into 1000 units. The knitting needle motion trajectory was set to Asin(ωt), A=5mm, ω=2π600rpm. The Maxwell-Wiechert model contains 3 Maxwell units, elastic modulus E1=1000MPa, E2=800MPa, E3=600MPa, viscosity coefficient η1=10Pa·s, η2=20Pa·s, η3=30Pa·s, and the dye adsorption capacity was introduced as a correction factor for the viscosity coefficient. The central difference method was used to solve the problem, with a time step of 1e-6 seconds and a total of 1,000,000 steps. The orthogonal experimental design L25(56) included six factors: dye molecular size, polarity, adsorption capacity, friction coefficient, yarn linear density, and weaving speed. The ARMA model order (2,1) processes time series data and calculates the tension fluctuation coefficient σ / μ, ranging from 2% to 10%.Multiple Linear Regression R. 2 =0.92, friction coefficient and dye adsorption capacity have the greatest impact on stability, with standardized regression coefficients of 0.45 and 0.38 respectively. Sensitivity analysis shows that for every 0.1 increase in friction coefficient, tension fluctuation increases by 1.2%. The experimental verification error is 4.3%, which is within the acceptable range.

[0046] Step S107, collect dye property yarn friction coefficient data in real time during the weaving process, compare the simulation results of the digital twin model, dynamically adjust the knitting needle motion trajectory equation and weaving parameters, and realize real-time optimization control of the weaving process.

[0047] The invention relates to a method for obtaining yarn friction coefficient data collected by an optical fiber sensor installed on a weaving machine; performing real-time simulation using a physics-based finite element digital twin model according to the yarn friction coefficient data and dye property data to obtain theoretical weaving parameters and needle motion trajectory; using a root mean square error method to compare the difference between the yarn friction coefficient data and the theoretical weaving parameters, and generating a deviation report including the error size and trend; according to the deviation report, constructing a weaving parameter optimization model using a deep Q learning network, and the weaving parameter optimization model is used to dynamically adjust the needle motion trajectory equation and weaving parameters; if the output result of the weaving parameter optimization model meets the preset performance indicator threshold, the sliding time window method is used to continuously evaluate the optimization effect of the weaving parameter optimization model; according to the optimization effect, dynamically adjusting the reward function of the deep Q learning network, and the reward function is used to achieve continuous optimization and closed-loop control of the weaving process.

[0048] Specifically, a fiber optic sensor is installed on the weaving machine to collect the data of the friction coefficient of the spinning yarn in real time, 100 times per second, and the dye property data, including the molecular size, polarity and adsorption capacity of the dye, are regularly obtained from the database. The collected data is filtered and standardized by the Kalman filter. A physical-based finite element digital twin model is used for real-time simulation, which is updated every 0.1 second. The processed dye property and friction coefficient data are input into the model, and the theoretical weaving parameters and needle motion trajectory are calculated to obtain the simulation results. The root mean square error method is used to compare the difference between the actual collected data and the simulation results, calculate the error value and deviation direction, and generate a deviation report containing the error size and trend. According to the deviation report, a weaving parameter optimization model is constructed using a deep Q learning network, and the needle motion trajectory equation and weaving parameters, including the needle lifting speed, yarn tension and yarn feeding speed, are dynamically adjusted, and the optimized parameters are fed back to the weaving machine control system in real time. The thresholds of key performance indicators are set, such as the yarn breakage rate is less than 0.1% and the fabric density deviation is less than 1%. The sliding time window method is used to continuously evaluate the optimization effect. The window size is 1 hour and updated every 10 minutes. According to the long-term performance, the reward function of the deep Q-learning network is dynamically adjusted to achieve continuous optimization and closed-loop control of the weaving process.

[0049] FBG fiber optic sensors with a sensitivity of 0.1pm / με were installed on the weaving machine to collect data on the friction coefficient of the spinning thread in real time with a sampling rate of 100Hz. The database updates the properties of the dye every hour, such as molecular size 3.5nm, polarity 2.8D, and adsorption energy -35kJ / mol. A 5th-order Kalman filter with a measurement noise covariance of R = 0.01 and a process noise covariance of Q = 1e-4 was used to filter the data. The finite element digital twin model based on ANSYS, containing 2400 BEAM188 units, was updated every 0.1 seconds to calculate the theoretical knitting needle motion trajectory Asin(ωt), A = 5mm, ω = 2π600rpm. The root mean square error was used to compare the actual data with the simulation results, such as the friction coefficient RMSE = 0.015, and a deviation report was generated. The deep Q learning network contains 3 fully connected layers, each with 128 neurons, a learning rate of 0.001, a discount factor γ = 0.99, an ε-greedy strategy ε = 0.1, and the target network is updated every 100 steps. Dynamically adjust the needle lifting speed ±5%, yarn tension ±2cN, and yarn feeding speed ±1m / min. The key performance indicators are set as yarn breakage rate <0.1% and fabric density deviation <1%. The sliding time window size is 1 hour, updated every 10 minutes, and the average performance indicators are calculated. Based on long-term performance, the reward function is adjusted every 24 hours, such as adjusting the fabric density deviation weight from 0.5 to 0.6. Through this closed-loop optimization process, the adjustment accuracy of weaving parameters has been improved by 15% and the stability of fabric quality has been improved by 8%.

[0050] Step S108: When a new dye formula is used in the knitting production line, the yarn friction coefficient is predicted according to the dye formula and the friction coefficient mapping library, and the weaving process is simulated and optimized through the digital twin model to provide the weaving parameter setting range and weaving process optimization suggestions.

[0051] The chemical composition and property parameters of the new dye formula in the dye formula database are obtained, and the property parameters include molecular weight, number of polar groups and spatial configuration; according to the chemical composition and property parameters, the weighted average method is used to calculate the comprehensive impact index. The comprehensive impact index is input into the pre-established random forest regression model, and the random forest regression model is used to predict the range of yarn friction coefficient. The predicted friction coefficient range is input into the digital twin model, and a multi-organization structural parameter combination is generated through Latin hypercube sampling. The multi-organization structural parameter combination includes needle movement speed, yarn tension and yarn feeding speed. According to the weaving parameter combination, the simulation experiment is run to simulate the weaving process; the simulation experiment results are processed using Python's pandas library to obtain the weaving efficiency, fabric quality index and energy consumption level under each set of parameter combinations. NSGA-II multi-objective optimization is performed on weaving efficiency, fabric quality indicators and energy consumption levels to screen out the optimal parameter combination that meets production requirements. A weaving parameter setting range report and process optimization suggestions are generated based on the optimal parameter combination. The process optimization suggestions include the value ranges of amplitude A, angular frequency ω and phase φ in the knitting needle motion trajectory equation S(t), the tension control curve represented by a piecewise linear function, and the yarn feeding speed adjustment strategy.

[0052] Specifically, the chemical composition and property parameters of the new dye formula, including molecular weight, number of polar groups and spatial configuration, are obtained from the dye formula database. The weighted average method is used to calculate the comprehensive impact considering the proportion of each component. The random forest regression model in the dye formula and friction coefficient mapping library is used to predict the range of yarn friction coefficient. The prediction model is adjusted through the transfer learning method to adapt to the characteristics of the new formula. The predicted friction coefficient range is input into the digital twin model. Combined with the characteristics of yarn linear density, strength and elongation, as well as the fabric structure and density parameters, Latin hypercube sampling is used to generate multi-tissue structural parameter combinations, including needle movement speed, yarn tension and yarn feeding speed. Run simulation experiments to simulate the weaving process. The pandas library of Python is used to process the simulation results, and the weaving efficiency, fabric quality index and energy consumption level under each set of parameter combinations are calculated. The NSGA-II algorithm is used for multi-objective optimization to screen out the optimal parameter combination that meets production requirements. According to the optimization results, a weaving parameter setting range report and process optimization suggestions are generated, including the amplitude A, angular frequency ω and phase φ value range in the needle motion trajectory equation S(t) = A*sin(ωt+φ), the tension control curve represented by the piecewise linear function, and the yarn feeding speed adjustment strategy. At the same time, the expected fabric performance indicators and production efficiency evaluation are provided. The new dye formula contains a molecular weight of 600g / mol, 8 polar groups, and a main component of planar spatial configuration, accounting for 70%. The weighted average calculation results in a comprehensive molecular weight of 520g / mol and 6.5 polar groups. The random forest regression model (100 decision trees, maximum depth 10) predicts a friction coefficient range of 0.25-0.35. Transfer learning uses fine-tuning of the last two layers, a learning rate of 0.001, and 50 iterations, and the prediction error is reduced from 8% to 3%. The digital twin model inputs yarn characteristics, linear density 30tex, strength 4cN / dtex, elongation 12%; fabric parameters, plain weave, density 250 strands / 10cm. Latin hypercube sampling generates 100 sets of parameter combinations, with a needle speed range of 500-700rpm, a tension range of 8-12cN, and a yarn feeding speed of 0.8-1.2m / s. ANSYS simulates 1000 needle cycles with a time step of 0.001s. Pandas processes the results and calculates weaving efficiency of 93-98%, coefficient of variation of fabric strength of 2.5-4%, and energy consumption of 0.15-0.18kWh / kg. The NSGA-II algorithm has a population size of 50, iterates for 100 generations, and selects 10 sets of optimal parameters. The optimization suggestion is that the needle trajectory S(t) = 5mm*sin(2π*600t+π / 6), the tension control curve 0-50% cycle 8cN, 50-100% cycle linear increase to 10cN, and the yarn feeding speed 1.0m / s±5% dynamic adjustment. The expected fabric strength CV <3%, production efficiency >95%.

[0053] Step S109, continuously collect data on dye properties, yarn friction coefficient and weaving process parameters, iteratively update and adaptively adjust the digital twin model and weaving process optimization system, improve the stability of weaving quality and the process adaptability of new dye formulas, and match the dye formula with weaving parameters.

[0054] The method comprises the following steps: obtaining yarn friction coefficient and weaving process parameter data collected by a high-precision sensor network, wherein the high-precision sensor network is installed on a weaving production line; detecting abnormal values ​​by a 3-times standard deviation method based on the yarn friction coefficient and weaving process parameter data, and filtering by a sliding window averaging method to obtain filtered parameter data; updating a digital twin model by an online stochastic gradient descent algorithm based on the filtered parameter data, and dynamically adjusting model parameters of the digital twin model; adjusting the weaving process in real time based on the digital twin model, and determining adjustment values ​​of needle movement speed, yarn tension and yarn feeding speed; obtaining association data between dye formula and weaving parameters, wherein the association data includes dye properties, friction coefficient and weaving parameters; analyzing the association data by an Apriori algorithm to obtain an intrinsic relationship between dye properties, friction coefficient and weaving parameters; generating a matching rule library based on the intrinsic relationship, and the matching rule library is used to dynamically adjust the control parameter weights.

[0055] Specifically, a high-precision sensor network is installed on the weaving production line to collect yarn friction coefficient and weaving process parameter data in real time. Dye property data is collected once for each batch, including dye molecular structure, polarity, yarn surface friction coefficient, needle motion trajectory, yarn tension and yarn feeding speed. Preliminary processing is performed through edge computing equipment, and the 3-times standard deviation method is used to detect outliers, and the sliding window average method is used for filtering. The digital twin model is updated using the online stochastic gradient descent algorithm, and the model parameters are dynamically adjusted to optimize the mapping relationship between dye properties and friction coefficient. The model predictive control algorithm is used to adjust the weaving process in real time. According to the actual production data and quality inspection results, the needle motion speed, yarn tension and yarn feeding speed are dynamically optimized to achieve closed-loop control of the weaving process. A database of dye formulas and weaving parameters is constructed, and the intrinsic relationship between dye properties, friction coefficients and weaving parameters is analyzed using the Apriori algorithm to generate a matching rule base. The mining results are applied to the model predictive control algorithm to dynamically adjust the control parameter weights. Regularly evaluate the overall performance of the system, including model prediction accuracy, fabric strength coefficient of variation, and process adaptability score of new dye formulas, and generate performance reports to provide a basis for further optimization. 100 high-precision fiber optic sensors are installed on the weaving production line with a sampling rate of 1000Hz to collect yarn friction coefficient and weaving parameters in real time. Dye property data is collected once every batch, about 8 hours, including molecular weight, range 200-800g / mol, polarity, 0-5D, and spatial configuration. The edge computing device uses the 3x standard deviation method to detect outliers, with a rejection rate of about 0.5%, and uses a 60-second sliding window average method for filtering. The online stochastic gradient descent algorithm updates the digital twin model once an hour, with a learning rate set to 0.01 and a mini-batch size of 128. The model prediction accuracy is improved from the initial 85% to 95%. The model predictive control algorithm uses a 15-minute prediction time domain and a 5-second control interval. The optimization objective function includes the fabric strength coefficient of variation, weighted 0.6, and energy consumption, weighted 0.4. The Apriori algorithm sets the minimum support of 0.05 and the minimum confidence of 0.7, and mines 500 valid rules from 100,000 production records. The rules are converted into control weights. For example, for every 1D increase in dye polarity, the yarn tension weight increases by 5%. The system generates a performance report once a week, including the root mean square error of model prediction, which is reduced to 0.02, the coefficient of variation of fabric strength, which is reduced from 4% to 2.5%, and the process adaptability score of the new dye formula, which is increased from 70 points to 85 points, with a full score of 100. Through continuous optimization, the stability of weaving quality has increased by 20%, and the adaptation cycle of new dye formulas has been shortened by 40%.

[0056] 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 large model-assisted knitting production line dynamic configuration and rapid response method, characterized in that: The method comprises: Based on the dye molecular weight, the number of dye polar groups, and the dye molecular spatial structure as dye property parameters, the historical data of the knitting production line is obtained and a dye property data warehouse is constructed; Based on the historical data of knitting production lines, dye molecular structure parameters, yarn surface morphology, material properties and friction coefficient test data were selected to construct a correlation prediction model between dye properties and yarn friction coefficient; The correlation data between dye properties and yarn friction coefficients were mined, and the correlation between dye molecular structure parameters and yarn surface morphology, material properties and friction coefficient numerical ranges was analyzed to obtain the distribution characteristics of yarn dye properties and friction coefficients under different dye property combinations; According to the dye properties and friction coefficient distribution characteristics, the dye formula composition is clustered and grouped, and the dye combination with similar influence on the yarn friction coefficient is determined to form a dye formula and friction coefficient mapping library; Acquire knitting machine structure parameters, needle motion trajectory equations, fabric structure matrix and fabric density parameters, and combine them with dye formula and friction coefficient mapping library to build a digital twin model of the weaving process; The dye properties, yarn characteristics and friction coefficient boundary conditions are set in the digital twin model, and the finite element simulation method is used to simulate the yarn movement and fabric formation process during weaving, and the influence of the dye properties and the spinning friction coefficient during weaving on the weaving process is analyzed; During the weaving process, the data of the friction coefficient of the dye properties and yarn are collected in real time, and the simulation results of the digital twin model are compared to dynamically adjust the needle motion trajectory equation and weaving parameters to achieve real-time optimization control of the weaving process; When a new dye formula is used in a knitting production line, the yarn friction coefficient is predicted based on the dye formula and the friction coefficient mapping library, and the weaving process is simulated and optimized through the digital twin model to provide weaving parameter setting ranges and weaving process optimization suggestions; Continuously collect data on dye properties, yarn friction coefficient and weaving process parameters, iteratively update and adaptively adjust the digital twin model and weaving process optimization system, improve the stability of weaving quality and the process adaptability of new dye formulas, and match dye formulas with weaving parameters.

2. A large model-assisted knitting production line dynamic configuration and rapid response method according to claim 1, characterized in that: The method uses dye molecular weight, dye polar group quantity, and dye molecular spatial structure as dye property parameters to obtain knitting production line historical data and build a dye property data warehouse, including: Obtaining dye molecular weight information and polar group quantity, wherein the dye molecular weight information is obtained from PubChem database, and the polar group quantity is calculated by SMILES string analysis method; According to the molecular weight information of the dye and the number of polar groups, a three-dimensional spatial structure model of the dye molecule is constructed using the PyMOL molecular modeling tool; Extracting the production batch information in the past five years from the production line history database, using Python's Pandas library to perform outlier detection and processing on the production batch information to obtain a standardized data set; Apache Hive was used to build a data warehouse, and the dye property parameters and standardized data sets were associated and mapped to fact tables and dimension tables, and multidimensional data cubes were built through Microsoft Analysis Services. The relationship between dye properties and production parameters was analyzed using correlation analysis methods, and a dye selection prediction model was constructed using the CART decision tree algorithm; If it is necessary to generate a dye selection recommendation list, the Apriori association rule mining algorithm is used to extract the association rules between dye properties and production parameters from the data warehouse.

3. A large model-assisted knitting production line dynamic configuration and rapid response method according to claim 1, characterized in that: The method comprises selecting dye molecular structure parameters, yarn surface morphology, material properties and friction coefficient test data based on the historical data of the knitting production line to construct a correlation prediction model between dye properties and yarn friction coefficient, including: Acquiring dye molecular structure parameters, wherein the dye molecular structure parameters include molecular weight, number of polar groups and spatial configuration; Analyze microscopic photos of the yarn surface to obtain fiber arrangement and surface roughness, and collect yarn material property data and friction coefficient; The raw data is preprocessed. If outliers are detected, they are removed. The spline interpolation algorithm is used to fill in the missing data. The data from different sources are formatted and converted to obtain processed data. Construct a multidimensional data model, where dye batches and yarn batches are used as fact tables, and dye molecular structure parameters, yarn surface morphology, material properties, and friction coefficient data are used as dimension tables; The time series analysis of historical data was performed to obtain the dynamic changes in the relationship between dye properties and yarn friction coefficient; Perform dimensionality reduction processing, calculate eigenvalues, select the principal component whose cumulative contribution rate reaches the preset threshold, and obtain the data after dimensionality reduction; Construct a correlation prediction model between dye properties and yarn friction coefficient and set the number of decision trees; The Gini index was used for feature selection, and the model accuracy was evaluated through cross-validation to obtain optimization suggestions for dye selection and yarn friction coefficient.

4. The large model-assisted knitting production line dynamic configuration and rapid response method according to claim 1 is characterized in that: The data on the correlation between dye properties and yarn friction coefficients are mined, and the correlation between dye molecular structure parameters and yarn surface morphology, material properties and friction coefficient numerical ranges is analyzed to obtain the dye properties and friction coefficient distribution characteristics of the yarn under different dye property combinations, including: Obtain dye molecular structure parameters, yarn surface morphology, material properties and friction coefficient data from the data warehouse; According to the data of dye molecular structure parameters, yarn surface morphology, material properties and friction coefficient, the Apriori algorithm is used to mine association rules, and a set of strong association rules between dye molecular structure parameters and yarn surface morphology, material properties and friction coefficient is obtained. According to the set of strong association rules, the dye property combinations are classified using a K-means clustering algorithm to determine the statistical distribution parameters of the yarn friction coefficient in each category; If the statistical distribution parameters meet the preset conditions, a multivariate linear regression method is used to establish a mathematical model between the dye properties and the yarn friction coefficient; Through mathematical models, the friction coefficient distribution characteristics of yarns under different combinations of dye properties are obtained; According to the distribution characteristics of friction coefficient, box plot and kernel density estimation method are used to generate visualization results; Determine whether the visualization result meets the preset threshold. If so, output the dye property and friction coefficient distribution feature report; It also includes: analyzing the correlation between the dye molecular structure, chromatographic characteristics and the yarn surface roughness and linear density value, and obtaining the surface morphological characteristics of the yarn under different dye properties.

5. The large model-assisted knitting production line dynamic configuration and rapid response method according to claim 1 is characterized in that: The method clusters and groups the dye formula compositions according to the dye properties and the friction coefficient distribution characteristics, determines the dye combinations with similar effects on the yarn friction coefficient, and forms a dye formula and friction coefficient mapping library, including: The dye property data and yarn friction coefficient data were obtained, and the Z-score standardization method was used to convert the data of different dimensions into a unified scale; Principal component analysis was performed on the standardized data to determine the principal component with a cumulative contribution rate of 95% as the input feature for cluster analysis; The DBSCAN algorithm is used to preprocess the input features to obtain a data set with outliers removed; The dye formula composition was grouped using the K-means clustering algorithm based on the data set, and the silhouette coefficient method was used to determine the optimal number of clusters; If the optimal number of clusters is determined, the dye property parameters and the corresponding friction coefficient mean of each cluster center are calculated; According to the clustering results, the mapping relationship between dye formula and friction coefficient is constructed; The CART decision tree algorithm was used to establish a prediction model from dye properties to friction coefficient, and the prediction model parameters were optimized through grid search. Generate a dye formula and friction coefficient mapping library according to the prediction model, the mapping library including dye combination features and corresponding friction coefficient prediction values; The sliding window method is used to update the mapping relationship between dye formula and friction coefficient in the mapping library.

6. A large model-assisted knitting production line dynamic configuration and rapid response method according to claim 1, characterized in that: The acquisition of knitting mechanical structure parameters, knitting needle motion trajectory equations, fabric structure matrix and fabric density parameters, combined with dye formula and friction coefficient mapping library, to construct a digital twin model of the weaving process includes: Acquire mechanical structure parameters in a knitting machine database, wherein the mechanical structure parameters include a needle cylinder diameter, a needle number, and a coil length; A high-precision position sensor is used to collect the knitting needle movement position information, and the knitting needle movement trajectory equation is obtained by fitting the position information through the cubic spline interpolation method; According to the fabric design drawings, convolutional neural networks are used to perform image segmentation and feature extraction to obtain the fabric structure matrix; Obtain fabric density parameters through a density measuring instrument, and store the tissue structure matrix and density parameters in a database; The association model between dye formula and friction coefficient and weaving parameters was established using random forest regression method; The dye recipes and data from the friction coefficient mapping library were matched with the weaving parameters to obtain a comprehensive data set; A long short-term memory network is used to construct a digital twin model of the weaving process, wherein the digital twin model takes mechanical parameters, trajectory equations, tissue structure matrix, density parameters and dye formula as inputs and friction coefficient as output; Optimize the parameters of the digital twin model through the back-propagation algorithm; Use 3D rendering technology to visualize the operation of the digital twin model, showing the movement of knitting needles and the formation of fabrics in real time; The digital twin model is optimized through the particle swarm optimization algorithm to improve the prediction accuracy.

7. A large model-assisted knitting production line dynamic configuration and rapid response method according to claim 1, characterized in that: The dye properties, yarn characteristics and friction coefficient boundary conditions are set in the digital twin model, the finite element simulation method is used to simulate the yarn movement and fabric formation process during weaving, and the influence of the dye properties and the spinning friction coefficient during weaving on the weaving process is analyzed, including: Acquire dye properties, yarn characteristics and friction coefficient data, and set boundary condition ranges, wherein the boundary condition ranges include dye molecular weight upper limit, dye molecular weight lower limit, polar group number upper limit, polar group number lower limit, yarn linear density upper limit, yarn linear density lower limit, yarn strength upper limit, yarn strength lower limit, yarn elongation upper limit, yarn elongation lower limit, friction coefficient upper limit, and friction coefficient lower limit; The yarn geometry model and the fabric geometry model are constructed using ANSYS finite element analysis software. If the yarn geometry model is constructed, the yarn geometry model is discretized into beam units, where each yarn is divided into a preset number of units. Determine the yarn material properties and contact conditions according to the yarn geometry model, and set the knitting needle motion trajectory and weaving tension as external loads; The weaving process is simulated using an explicit dynamics solver, which determines the time step based on the Courant-Friedrichs-Lewy condition; An orthogonal experimental design method is used to set up an experimental scheme, wherein the experimental scheme includes the number of factors and the number of levels, a simulation experiment is performed according to the experimental scheme, and a significance test is performed on the simulation results using variance analysis; ParaView scientific visualization software was used to display the yarn motion trajectory and stress distribution, and the weight model of influencing factors was established through grey correlation analysis. The weight model of influencing factors was used to quantitatively describe the influence of each factor on the weaving process. It also includes: setting the dye molecular size, polarity, adsorption capacity property parameters and yarn friction coefficient as boundary conditions, using finite element simulation to simulate the dynamic changes of yarn tension, speed and needle motion trajectory during weaving, and analyzing the influence of dye properties and friction coefficient on the stability of the weaving process.

8. A large model-assisted knitting production line dynamic configuration and rapid response method according to claim 7, characterized in that: The method sets the dye molecular size, polarity, adsorption capacity property parameters and yarn friction coefficient as boundary conditions, uses finite element simulation to simulate the dynamic changes of yarn tension, speed and knitting needle motion trajectory during weaving, and analyzes the influence of dye properties and friction coefficient on the stability of the weaving process, including: Acquiring dye molecular size, polarity and adsorption capacity data, the data being obtained from a pre-established dye database; measuring the yarn friction coefficient using an inclined plane method, the measurement being performed under preset temperature and humidity conditions; According to the acquired dye data and friction coefficient, a numerical model of the weaving process is constructed, wherein the yarn is discretized into predefined beam units and the knitting needle motion trajectory equation is set; Use explicit dynamics solver to simulate the weaving process; calculate the tension, velocity and displacement of the yarn at each time step, and record the real-time coordinates of the needle motion trajectory; Through the orthogonal experimental design method, a multi-factor and multi-level experimental scheme was set up; multiple simulation experiments were carried out to calculate the fluctuation coefficient of yarn tension, speed and knitting needle motion trajectory; If the fluctuation coefficient is less than the preset threshold, the weaving process is judged to be stable; otherwise, a sensitivity analysis is performed to determine the influence of each parameter on the stability.

9. A large model-assisted knitting production line dynamic configuration and rapid response method according to claim 1, characterized in that: The real-time collection of dye property spinning friction coefficient data during the weaving process, comparison of the simulation results of the digital twin model, dynamic adjustment of the knitting needle motion trajectory equation and weaving parameters, and the realization of real-time optimization control of the weaving process include: Obtaining yarn friction coefficient data collected by an optical fiber sensor, wherein the optical fiber sensor is installed on a weaving machine; According to the data of yarn friction coefficient and dye property, the physics-based finite element digital twin model is used for real-time simulation to obtain the theoretical weaving parameters and needle motion trajectory; The root mean square error method is used to compare the difference between the yarn friction coefficient data and the theoretical weaving parameters, and a deviation report including the error size and trend is generated; According to the deviation report, a weaving parameter optimization model is constructed using a deep Q learning network, and the weaving parameter optimization model is used to dynamically adjust the knitting needle motion trajectory equation and weaving parameters; If the output result of the weaving parameter optimization model meets the preset performance index threshold, the optimization effect of the weaving parameter optimization model is continuously evaluated using the sliding time window method; According to the optimization effect, the reward function of the deep Q learning network is dynamically adjusted, and the reward function is used to achieve continuous optimization and closed-loop control of the weaving process.

10. The large model-assisted knitting production line dynamic configuration and rapid response method according to claim 1, characterized in that: When a new dye formula is used in the knitting production line, the yarn friction coefficient is predicted according to the dye formula and the friction coefficient mapping library, and the weaving process is simulated and optimized through the digital twin model, and the weaving parameter setting range and weaving process optimization suggestions are given, including: Obtaining the chemical composition and property parameters of a new dye formula in a dye formula database, wherein the property parameters include molecular weight, number of polar groups and spatial configuration; According to the chemical composition and property parameters, the comprehensive impact index is calculated by weighted average method; Inputting the comprehensive impact index into a pre-established random forest regression model, the random forest regression model is used to predict the range of yarn friction coefficient; The predicted friction coefficient range is input into the digital twin model, and a multi-tissue structural parameter combination is generated through Latin hypercube sampling. The multi-tissue structural parameter combination includes needle movement speed, yarn tension and yarn feeding speed. Run simulation experiments based on the weaving parameter combinations to simulate the weaving process; The simulation results were processed using Python's pandas library to obtain the weaving efficiency, fabric quality index, and energy consumption level under each parameter combination. NSGA-II multi-objective optimization was performed on weaving efficiency, fabric quality index and energy consumption level to screen out the optimal parameter combination that meets production requirements; A weaving parameter setting range report and process optimization suggestions are generated based on the optimal parameter combination. The process optimization suggestions include the value ranges of amplitude A, angular frequency ω and phase φ in the needle motion trajectory equation S(t), the tension control curve represented by a piecewise linear function, and the yarn feeding speed adjustment strategy.

Images