Process optimization method for coloring artificial fur fabric by using cations

Abstract

The invention provides a process optimization method for coloring an artificial fur fabric by using cations, which belongs to the technical field of artificial fur fabric coloring processes, and comprises the following steps: collecting various process parameters and quality index data required by artificial fur fabric coloring, including grey cloth specification, squeezing treatment, dye formula, dyeing process, post-treatment and the like. Establishing a coloring effect empirical model considering each process parameter, wherein the coloring effect empirical model comprises equations of dye diffusion, fixation, fiber water absorption, fluff shaping and the like; substituting the collected parameter data into the model to obtain a preliminary process parameter-effect relationship; a large number of technological parameter combinations are generated according to the parameter range, and clustering analysis is carried out by utilizing a model calculation theoretical effect. And selecting process parameters of the typical clustering center to carry out pilot test verification, and optimizing the process parameter-effect relation model. And adopting a genetic algorithm to obtain an optimal process parameter combination on the basis of the optimization model, carrying out pilot scale production verification, and finally determining an optimization process scheme.

Description

Technical Field

[0001] The invention belongs to the technical field of artificial fur fabric coloring technology, and in particular relates to a process optimization method for coloring artificial fur fabric using cations. Background Art

[0002] Artificial fur fabrics are widely used in clothing, home textiles, toys and other fields because they look like natural fur, are relatively cheap and do not involve animal protection issues. As consumers' demands for product quality and personalization continue to increase, the coloring process of artificial fur fabrics has become a key factor in product quality and market competitiveness. The traditional coloring process of artificial fur fabrics mainly relies on experience accumulation and repeated experiments, and there are problems such as long process parameter optimization cycle, high resource consumption, and unstable product quality.

[0003] At present, the artificial fur fabric coloring process commonly used in the industry mainly includes the steps of grey cloth preparation, water squeezing treatment, dye preparation, dyeing, water washing softener treatment, drying, shaping, etc. In the dyeing process, disperse dyes or acid dyes are usually used, but these dyes have poor affinity for man-made fibers, resulting in poor color fastness, low color saturation and other problems. In order to improve the dyeing effect, some manufacturers try to use cationic dyes, but due to the lack of a systematic process parameter optimization method, it is often necessary to determine the optimal process conditions through a large number of trial and error experiments, which is not only time-consuming and labor-intensive, but also difficult to ensure the consistency of product quality. Summary of the invention

[0004] In view of this, the present invention provides a process optimization method for coloring artificial fur fabric using cationic materials, which can solve the technical problems of the existing process optimization methods that are time-consuming and labor-intensive.

[0005] The present invention is achieved in that:

[0006] The present invention provides a process optimization method for coloring artificial fur fabric using cationic materials, which comprises the following steps:

[0007] S10, collecting various parameter data of the artificial fur fabric coloring process, including grey fabric specifications, squeezing treatment parameters, dye formula, dyeing process parameters, post-treatment process parameters and final product quality data;

[0008] S20, establishing a group of empirical equations for coloring effects taking into account various process parameters;

[0009] S30, fitting the coloring effect empirical equation group using the collected parameter data to obtain a preliminary process parameter-effect relationship model;

[0010] S40, generating a large number of process parameter combination data sets in a traversal manner according to a preset process parameter range and step size, and using the process parameter-effect relationship preliminary model to perform calculations to obtain a large number of candidate process parameter combinations and their theoretical effect data;

[0011] S50, performing cluster analysis on a large number of candidate process parameter combinations according to theoretical effect data to obtain multiple cluster centers;

[0012] S60, selecting a process parameter combination corresponding to each cluster center as a process set to be tested, conducting a small-scale production test, and collecting actual effect data;

[0013] S70, comparing the actual effect data with the theoretical effect data, optimizing the preliminary model of the process parameter-effect relationship, and obtaining a process parameter-effect relationship optimization model;

[0014] S80, taking multiple groups of process parameters with the best actual effects in the process to be tested as the initial population, setting the crossover rate and mutation rate, and using a genetic algorithm to obtain the optimal process parameter combination through multiple generations of iterative optimization;

[0015] S90. Use the optimal process parameter combination to carry out pilot production, verify the optimization effect, and fine-tune the process parameters according to the pilot results, and finally determine the optimized process plan for coloring artificial fur fabrics.

[0016] Specifically, the steps S10 and S20 specifically include:

[0017] Step 101, collecting various parameter data of the artificial fur fabric coloring process, including grey fabric specifications, water squeezing treatment parameters, dye formula, dyeing process parameters, post-processing process parameters and final product quality data;

[0018] Step 102, sorting out the specification data of the grey fabric, including the raw material composition, pile height, width and gram weight of the weft knitted artificial fur;

[0019] Step 103, recording water-squeezing processing parameters, including water temperature, vehicle speed and pressure;

[0020] Step 104: sorting out dye formula information, including types and ratios of cationic dyes and high-temperature disperse dyes;

[0021] Step 105, collecting dyeing process parameters, including dyeing temperature, time and bath ratio;

[0022] Step 106, summarizing the post-processing process parameters, including specific parameters of washing and softening agent treatment, drying, base fabric sizing and shaping, brushing, ironing, hot blowing and shaping;

[0023] Step 107, collecting final product quality data, including color fastness, hand feel, fuzz fastness and color uniformity;

[0024] Step 108, establishing a dye diffusion equation, taking into account the changes in the concentration distribution of the dye in the fiber over time and space;

[0025] Step 109, construct a dye fixation equation to describe the chemical bonding process between the dye and the fiber, taking into account the effects of temperature, pH value and auxiliaries;

[0026] Step 110, establishing a fiber water absorption equation to describe the water absorption behavior of the fiber during the dyeing process, taking into account the influence of capillary action and fiber structure;

[0027] Step 111, constructing a fluff setting equation to describe the setting effect of the fluff during the heat treatment process, taking into account the influence of temperature, pressure and humidity;

[0028] Step 112, establishing a color saturation equation to relate dye concentration, fiber characteristics and color space parameters;

[0029] Step 113: Construct a correlation equation for the process steps to describe the mutual influence and cumulative effect between the various process steps.

[0030] Wherein, the step S30 and step S40 specifically include:

[0031] Step 301: preprocess the collected data, including data cleaning, normalization and standardization;

[0032] Step 302: Select a machine learning algorithm, including support vector machine regression, random forest regression, and gradient boosting tree;

[0033] Step 303: Divide the data set into a training set and a validation set, using a ratio of 80% to 20%;

[0034] Step 304: Use grid search and cross-validation methods to tune the model's hyperparameters;

[0035] Step 305: train the machine learning model, using the root mean square error and the coefficient of determination as evaluation indicators;

[0036] Step 306: Perform an interpretative analysis on the trained model using feature importance analysis and partial dependency graphs;

[0037] Step 401, determining the reasonable variation range and step length of each process parameter, including dyeing temperature, dyeing time and dye concentration;

[0038] Step 402: Generate all possible parameter combinations using Cartesian product algorithm;

[0039] Step 403: For each set of parameter combinations, use the trained machine learning model to make predictions to obtain theoretical effect data;

[0040] Step 404: Screen the prediction results to remove results that are obviously unreasonable or beyond an acceptable range;

[0041] Step 405: preliminarily sort the remaining candidate process parameter combinations and calculate the comprehensive scores using the weighted average method;

[0042] Step 406: Save the 1000 candidate process parameter combinations with the highest scores and their theoretical effect data.

[0043] Wherein, the step S50 and step S60 specifically include:

[0044] Step 501, standardize the theoretical effect data using the Z-score standardization method;

[0045] Step 502: Select the density clustering algorithm DBSCAN to perform cluster analysis;

[0046] Step 503: Setting the neighborhood radius and minimum number of points parameters of the DBSCAN algorithm;

[0047] Step 504: Run the clustering algorithm to obtain preliminary clustering results;

[0048] Step 505: Use the silhouette coefficient to evaluate the clustering quality, and optimize the clustering results by adjusting parameters;

[0049] Step 506: Perform feature analysis on each cluster and calculate the cluster center;

[0050] Step 507: Use principal component analysis to reduce the high-dimensional data to a two-dimensional or three-dimensional space for visualization;

[0051] Step 508: According to the clustering results, 5 to 10 representative process parameter combinations are selected as the process set to be tested;

[0052] Step 601: design a small-scale production test plan according to the selected process set to be tested;

[0053] Step 602: Prepare experimental materials and equipment to ensure the consistency of raw material batches and the stability of equipment status;

[0054] Step 603: Conduct small-scale production tests in sequence according to the designed process parameter combinations;

[0055] Step 604, recording real-time data during the production process, including temperature curve and pH value changes;

[0056] Step 605, conduct comprehensive quality tests on the produced samples, including color fastness, hand feel, nap firmness and color uniformity;

[0057] Step 606: Perform statistical analysis on the test data to calculate the average value and standard deviation of each indicator;

[0058] Step 607: Compare and analyze the actual effect data with the theoretical prediction data, and calculate the prediction error.

[0059] Wherein, the step S70 specifically includes:

[0060] Step 701: Pair the actual effect data with the theoretical prediction data, and calculate the absolute error and relative error;

[0061] Step 702: Analyze the error distribution characteristics and use box plots and scatter plots for visual display;

[0062] Step 703: Adjust and optimize the original model according to the error analysis results;

[0063] Step 704: Perform feature engineering to create new features, including square terms and cross terms of parameters;

[0064] Step 705: Adjust the model structure, such as adjusting the number of layers and neurons of the neural network, or adjusting the depth of the tree model and the minimum number of samples of leaf nodes;

[0065] Step 706: Apply an ensemble learning method to combine multiple basic models, using random forest or gradient boosting tree;

[0066] Step 707: Introduce L1 or L2 regularization term to prevent overfitting of the model;

[0067] Step 708: Using transfer learning, use a model pre-trained on a large amount of historical data and fine-tune it to adapt to the current task;

[0068] Step 709: retrain and evaluate the optimized model using a 5-fold or 10-fold cross validation method;

[0069] Step 710, comparing the performance of the model before and after optimization, the evaluation indicators include root mean square error, mean absolute error and determination coefficient;

[0070] Step 711: If the performance of the optimized model is significantly improved, the new model is adopted; otherwise, the cause is further analyzed and the optimization strategy is adjusted;

[0071] Step 712: Use the optimized model to re-predict the effect data of a large number of process parameter combinations and update the candidate process parameter set.

[0072] Wherein, the step S80 specifically includes:

[0073] Step 801, selecting 3 to 5 sets of process parameters with the best actual effects from the experimental results as the initial population;

[0074] Step 802: Design a fitness function and use a weighted average method to calculate the comprehensive effect score;

[0075] Step 803, setting the parameters of the genetic algorithm, including population size, crossover rate, mutation rate and number of iterations;

[0076] Step 804: Initialize the population, and in addition to the selected optimal combination, randomly generate other individuals to ensure population diversity;

[0077] Step 805: Evaluate the fitness, use the optimized model to predict the effect of each individual, and calculate the fitness value;

[0078] Step 806: Select outstanding individuals using the tournament selection method;

[0079] Step 807: Perform single-point crossover on the selected individuals to generate new offspring;

[0080] Step 808: Perform Gaussian mutation on some individuals to introduce randomness;

[0081] Step 809: Update the population, add the newly generated offspring to the population, and eliminate individuals with lower fitness;

[0082] Step 810, repeating steps 805 to 809 until the termination condition is met;

[0083] Step 811, performing feasibility analysis on the optimal process parameter combination finally obtained;

[0084] Step 812: Check whether the parameters are within the range allowed by the equipment and whether they are in line with actual production;

[0085] Step 813: If there are infeasible parameters, adjust them or re-run the algorithm;

[0086] Step 814: output the optimal process parameter combination and its predicted comprehensive effect score.

[0087] Wherein, the step S90 specifically includes:

[0088] Step 901: Develop a detailed pilot production plan based on the optimal process parameter combination;

[0089] Step 902: Conduct preparatory work before pilot production, including raw material inspection, equipment calibration and personnel training;

[0090] Step 903: Conduct pilot production according to the formulated plan and strictly control various process parameters;

[0091] Step 904: Real-time monitoring and recording of changes in key process parameters during the production process;

[0092] Step 905: sampling at key process nodes and testing intermediate products;

[0093] Step 906: Conduct comprehensive quality inspection on the pilot-produced products, including color fastness, hand feel, nap firmness, and color uniformity;

[0094] Step 907: adding tests simulating actual use, such as performance changes after repeated washing and wearing comfort evaluation;

[0095] Step 908: Compare and analyze the pilot test results with the small test results, and calculate the differences in various indicators;

[0096] Step 909, analyzing the causes of the differences, including differences in heat transfer and mass transfer caused by changes in production scale, differences in equipment accuracy, and differences in raw material batches;

[0097] Step 910: fine-tune the process parameters according to the analysis results;

[0098] Step 911: Use the orthogonal experiment method to design a fine-tuning experiment to quickly find the best parameter combination;

[0099] Step 912: Comprehensively consider factors such as product quality, production efficiency and cost to determine the final optimized process plan;

[0100] Step 913: Establish the allowable fluctuation range of process parameters to provide guidance for subsequent large-scale production.

[0101] The process of coloring the artificial fur fabric using cationic dyes includes the following steps: preparation of grey cloth, water squeezing treatment, preparation of cationic dyes, preparation of high-temperature disperse dyes, integrated dyeing of cationic high-temperature disperse dyes, water-washing softener treatment, dehydration, drying, sizing and shaping of the base fabric, brushing, ironing, hot blowing, shaping, and ironing finishing.

[0102] Among them, the quality data includes color fastness, feel, nap firmness and color uniformity.

[0103] Furthermore, the coloring effect empirical equation group includes a dye diffusion equation, a dye fixation equation, a fiber water absorption equation, a fluff setting equation, a color saturation equation, and correlation equations between various process steps.

[0104] The coloring effect empirical equation group is specifically described as follows:

[0105] 1. The dye diffusion equation is specifically expressed as follows:

[0106] ;

[0107] In the formula, is the concentration of dye in the fiber (g / L); is the staining time (s); is the dye diffusion coefficient (m² / s); is the spatial coordinate (m); is the maximum dye adsorption capacity of the fiber (g / L); is the dye adsorption rate constant (L / (g·s)); is the dye desorption rate constant (1 / s).

[0108] Parameter acquisition method:

[0109] It is obtained through dyeing kinetics experiment, the steps are as follows:

[0110] (1) Carry out dyeing experiments at different temperatures and measure the dye uptake rate at different time points;

[0111] (2) According to Fick's diffusion law, the diffusion coefficient at different temperatures is fitted;

[0112] (3) The relationship between the diffusion coefficient and temperature was obtained by fitting the Arrhenius equation.

[0113] It is obtained through isothermal adsorption experiment, the steps are as follows:

[0114] (1) Prepare dye solutions of different concentrations, add equal amounts of fibers, and adsorb to equilibrium at the equilibrium temperature;

[0115] (2) Determine the dye concentration in the solution after equilibrium and calculate the fiber adsorption amount;

[0116] (3) The maximum adsorption capacity was obtained by fitting the Langmuir isotherm adsorption equation.

[0117] and It is obtained through kinetic experiments, the steps are as follows:

[0118] (1) Carry out dyeing experiments at different temperatures and measure the dye uptake rate at different time points;

[0119] (2) The pseudo-second-order kinetic equation was used to fit the and ;

[0120] (3) Using the Arrhenius equation to fit and Relationship with temperature.

[0121] 2. The dye fixation equation is specifically expressed as follows:

[0122] ;

[0123] In the formula, is the dye fixation rate (%); is the maximum fixation rate (%); is the fixation rate constant (1 / s); is the fixation time (s); is the periodic fluctuation amplitude (%); is the fluctuation frequency (rad / s); is the phase angle (rad); is the first error term.

[0124] Parameter acquisition method:

[0125] It is obtained through fixation experiment, the steps are as follows:

[0126] (1) Carry out fixation experiments at different temperatures, pH values ​​and additive concentrations, and measure the fixation rates at different time points;

[0127] (2) Use nonlinear least squares method to fit the parameters;

[0128] (3) Analyze the relationship between each parameter and temperature, pH and additive concentration, and establish a multivariate regression model.

[0129] 3. The fiber water absorption equation is specifically expressed as follows:

[0130] ;

[0131] In the formula, is the fiber moisture content (%); is the initial moisture content (%); is the maximum water absorption (%); is the water absorption rate constant (1 / s); is the water absorption time (s); is the capillary action coefficient (% / s^(1 / 2)); is the periodic fluctuation amplitude (%); is the fluctuation frequency (rad / s); is the second error term.

[0132] Parameter acquisition method:

[0133] It is obtained through water absorption experiment, the steps are as follows:

[0134] (1) Determination of initial fiber moisture content ;

[0135] (2) Immerse the fiber in water and take it out at different time points to measure the moisture content;

[0136] (3) Use nonlinear least squares method to fit the parameters;

[0137] (4) Analyze the relationship between each parameter and fiber structure, surface treatment and other factors.

[0138] 4. The villi shaping equation is specifically expressed as follows:

[0139] ;

[0140] In the formula, is the fluff setting degree (%); is the maximum setting degree (%); is the stereotyped rate constant (1 / s); is the setting time (s); is the setting temperature (°C); is the setting pressure (kPa); is the relative humidity (%); is the influence coefficient of temperature, pressure and humidity on setting degree; is the third error term.

[0141] Parameter acquisition method:

[0142] It is obtained through stereotyping experiments, the steps are as follows:

[0143] (1) Carry out setting experiments under different temperature, pressure and humidity conditions and measure the setting degree at different time points;

[0144] (2) Use nonlinear least squares method to fit the parameters;

[0145] (3) Analyze the relationship between each parameter and factors such as fiber type and fabric structure, and establish a multivariate regression model.

[0146] 5. The color saturation equation is specifically expressed as follows:

[0147] ;

[0148] In the formula, is the color saturation; is the maximum saturation; is the influence coefficient of dye concentration on saturation; is the dye concentration (g / L); is the CIELAB color space coordinate; for The coefficient of influence on saturation; is the fourth error term.

[0149] Parameter acquisition method:

[0150] It is obtained through dyeing experiments and color measurement. The steps are as follows:

[0151] (1) Perform staining experiments using dyes of different concentrations;

[0152] (2) Measure the dyed sample value and saturation;

[0153] (3) Use nonlinear least squares method to fit the parameters;

[0154] (4) Analyze the relationship between each parameter and factors such as dye type and fiber structure.

[0155] 6. The process step correlation equation is specifically expressed as follows:

[0156] ;

[0157] In the formula, For the overall process effect; is the number of process steps; For steps and steps The interaction coefficient of ; For steps The effect function; For steps The process parameter vector; For steps The main effect coefficients of ; is the fifth error term.

[0158] Parameter acquisition method:

[0159] It is obtained through orthogonal experiment and variance analysis, the steps are as follows:

[0160] (1) Design an orthogonal experimental plan, including the key parameters of each process step;

[0161] (2) Conduct experiments and measure overall process effects;

[0162] (3) Calculate the main effects and interactions of each step by analysis of variance;

[0163] (4) Use multiple regression analysis to obtain the coefficients.

[0164] The functional form is determined by a single factor experiment, and the steps are as follows:

[0165] (1) Fix other step parameters and change the step Experiment with the parameters;

[0166] (2) measure the effectiveness of this step;

[0167] (3) Try different function forms (such as linear, quadratic, exponential, etc.) to fit the experimental data and select the function form with the best fitting effect.

[0168] Specifically, each function is determined in the following way:

[0169] 1. Water treatment effect function The specific expressions are as follows:

[0170] ;

[0171] In the formula, is water temperature (℃); is the vehicle speed (yards / minute); is the pressure (KG); is the coefficient to be determined; is the error term.

[0172] Parameter acquisition method:

[0173] (1) Design an orthogonal experiment, considering three factors: water temperature (40-50℃), vehicle speed (30-50 mph), and pressure (1.5-2.5 kg);

[0174] (2) Carry out water squeezing treatment experiment and measure the moisture content of the fabric after treatment;

[0175] (3) The coefficients are obtained by fitting using the nonlinear least squares method.

[0176] 2. Dyeing effect function The specific expressions are as follows:

[0177] ;

[0178] In the formula, is the staining time (min); is dyeing temperature (℃); is the dye concentration (g / L); is the pH value of the dye bath; is the coefficient to be determined; is the error term.

[0179] Parameter acquisition method:

[0180] (1) Design the Box-Behnken experiment, considering four factors: dyeing time (30-90 min), temperature (80-120 °C), dye concentration (1-5 g / L), and pH value (4-7);

[0181] (2) Conduct dyeing experiments and measure the K / S value to characterize the dyeing depth;

[0182] (3) The coefficients were obtained by fitting using the response surface methodology.

[0183] 3. Washing softener treatment effect function The specific expressions are as follows:

[0184] ;

[0185] In the formula, is the softener concentration (%); is the electrostatic agent concentration (%); is the vehicle speed (yards / minute); is the treatment temperature (℃); is the pressure (KG); is the coefficient to be determined; is the error term.

[0186] Parameter acquisition method:

[0187] (1) Design a central composite experiment, considering five factors: softener concentration (2-5%), electrostatic agent concentration (0.5-1.5%), vehicle speed (30-50 yards), temperature (40-50℃), and pressure (1.5-2.5KG);

[0188] (2) Carry out washing softener treatment experiment to measure fabric softness and electrostatic performance;

[0189] (3) Use multiple linear regression fitting to obtain the coefficients.

[0190] 4. Drying effect function The specific expressions are as follows:

[0191] ;

[0192] In the formula, is the drying time (min); is the drying temperature (℃); is the vehicle speed (yards / minute); is the oven humidity (%); is the temperature fluctuation frequency (rad / min); is the coefficient to be determined; is the error term.

[0193] Parameter acquisition method:

[0194] (1) Design an orthogonal experiment, considering four factors: drying time (5-15 min), temperature (140-180°C), vehicle speed (35-55 mph), and humidity (30-70%);

[0195] (2) Conduct drying experiments to measure the moisture content and feel of the fabric;

[0196] (3) The coefficients are obtained by fitting using the nonlinear least squares method.

[0197] 5. Function of base fabric sizing and shaping effect The specific expressions are as follows:

[0198] ;

[0199] In the formula, is the slurry concentration (%); is the setting temperature (℃); is vehicle speed (m / min); is wind speed (r / min); is the coefficient to be determined; is the error term.

[0200] Parameter acquisition method:

[0201] (1) Design a partial factorial experiment, considering four factors: slurry concentration (25-80%), temperature (170-200℃), vehicle speed (30-45m / min), and wind speed (600-1450rpm);

[0202] (2) Conduct base fabric sizing and shaping experiments to measure fabric stiffness and dimensional stability;

[0203] (3) Use multivariate nonlinear regression fitting to obtain the coefficients.

[0204] 6. Brush effect function The specific expressions are as follows:

[0205] ;

[0206] In the formula, is the vehicle speed (yards / minute); for tension; is the speed (rpm); is the depth of hair feeding (mm); is the coefficient to be determined; is the error term.

[0207] Parameter acquisition method:

[0208] (1) Design the Plackett-Burman experiment, considering four factors: vehicle speed (4-8 yds), ​​tension (1020-1032), rotation speed (1000-11000 rpm), and wool feeding depth (1.0-2.4 mm);

[0209] (2) Conduct brushing experiments to measure the density and uniformity of the hair;

[0210] (3) The coefficients were obtained by stepwise regression analysis.

[0211] Furthermore, the method of fitting the coloring effect empirical equation group using the collected parameter data is least square fitting.

[0212] Furthermore, the preset process parameter range and step size, wherein the step size is 0.01~0.1 of the corresponding process parameter range.

[0213] Furthermore, K-means clustering was used for cluster analysis.

[0214] Furthermore, the genetic algorithm uses the comprehensive effect score obtained by the process parameter-effect relationship optimization model as a fitness function.

[0215] The crossover rate and mutation rate of the genetic algorithm can be set based on experience or directly use the recommended values ​​or default values ​​of the genetic algorithm.

[0216] Furthermore, the comprehensive effect score is a weighted average of color fastness, feel, fuzz fastness and color uniformity.

[0217] Furthermore, the iteration stopping condition of the genetic algorithm includes reaching a maximum number of iteration steps of 2000 times.

[0218] Furthermore, the iteration stopping condition of the genetic algorithm is that the difference between the fitness function values ​​of two consecutive iteration results is less than 0.00375.

[0219] Compared with the prior art, the process optimization method for coloring artificial fur fabric using cationic materials provided by the present invention has the following beneficial effects:

[0220] First, the present invention establishes a set of empirical equations for the coloring effect taking into account various process parameters, which mathematically transforms the complex dyeing process and lays a theoretical foundation for subsequent optimization analysis. These equations include dye diffusion equations, dye fixation equations, fiber water absorption equations, etc., which fully consider the physical and chemical mechanisms in the dyeing process, making the optimization process no longer a "black box" operation, but having a clear theoretical basis.

[0221] Secondly, the present invention uses machine learning algorithms, especially support vector machine regression, random forest regression and gradient boosting tree methods, to model the process parameter-effect relationship. This method can not only handle high-dimensional parameter space, but also capture the complex nonlinear relationship between parameters, greatly improving the prediction accuracy of the model. Compared with traditional statistical methods, the method of the present invention shows stronger adaptability and robustness when dealing with complex systems.

[0222] Furthermore, the present invention generates a large number of process parameter combinations by traversal, and uses the trained model for rapid prediction. This method greatly expands the parameter search space and increases the probability of finding the optimal solution. At the same time, through cluster analysis, the present invention can effectively classify a large number of candidate solutions and find representative parameter combinations, which greatly reduces the number of actual experiments and saves time and resources.

[0223] In addition, the present invention collects actual effect data through small-scale production tests and compares it with theoretical prediction results to continuously optimize and improve the model. This closed-loop feedback mechanism ensures that the model can continue to learn and adapt to actual production conditions, improving the reliability and practicality of the optimization results.

[0224] Finally, the present invention uses a genetic algorithm for multi-generation iterative optimization, which can quickly converge to the global optimal solution in a complex parameter space. Different from the traditional single-objective optimization, the method of the present invention uses the comprehensive effect score as the fitness function to achieve multi-objective balanced optimization, which better meets the actual production needs.

[0225] In summary, the technical solution of the present invention has the following significant improvements compared with the prior art: 1) The optimization efficiency is greatly improved, shortening the process optimization process that traditionally takes months or even years to a few weeks; 2) The quality of the optimization results is higher, and multiple indicators such as color fastness, feel, fluff firmness and color uniformity can be taken into account at the same time; 3) Resource consumption is significantly reduced, reducing a large number of trial and error experiments; 4) The optimization process is more intelligent and automated, reducing dependence on manual experience; therefore, the solution of the present invention solves the technical problem that the existing process optimization methods are time-consuming and labor-intensive. BRIEF DESCRIPTION OF THE DRAWINGS

[0226] Figure 1 A flow chart of the method provided by the present invention;

[0227] Figure 2 It is a comparison diagram of the effects of the embodiments of the present invention. DETAILED DESCRIPTION

[0228] In order to make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0229] like Figure 1 As shown, it is a flow chart of a process optimization method for coloring artificial fur fabric using cationic materials provided by the present invention. The method comprises the following steps:

[0230] S10. Collect various parameter data of the artificial fur fabric coloring process, including grey fabric specifications, squeezing treatment parameters, dye formula, dyeing process parameters, post-treatment process parameters and final product quality data, wherein the quality data includes color fastness, hand feel, nap firmness and color uniformity;

[0231] S20, establishing a group of empirical equations for coloring effects taking into account various process parameters, including a dye diffusion equation, a dye fixation equation, a fiber water absorption equation, a fluff setting equation, a color saturation equation, and correlation equations between various process steps;

[0232] S30, fitting the coloring effect empirical equation group using the collected parameter data to obtain a preliminary process parameter-effect relationship model;

[0233] S40, generating a large number of process parameter combination data sets in a traversal manner according to a preset process parameter range and step size, and using the process parameter-effect relationship preliminary model to perform calculations to obtain a large number of candidate process parameter combinations and their theoretical effect data;

[0234] S50, performing cluster analysis on a large number of candidate process parameter combinations according to theoretical effect data to obtain multiple cluster centers;

[0235] S60, selecting a process parameter combination corresponding to each cluster center as a process set to be tested, conducting a small-scale production test, and collecting actual effect data;

[0236] S70, comparing the actual effect data with the theoretical effect data, optimizing the preliminary model of the process parameter-effect relationship, and obtaining a process parameter-effect relationship optimization model;

[0237] S80, using multiple groups of process parameters with the best actual effects in the process set to be tested as the initial population, using the comprehensive effect score obtained by the process parameter-effect relationship optimization model as the fitness function, setting the crossover rate and mutation rate, using a genetic algorithm, and obtaining the optimal process parameter combination through multi-generation iterative optimization, wherein the comprehensive effect score is a weighted average of color fastness, hand feel, fuzz fastness, and color uniformity;

[0238] S90. Use the optimal process parameter combination to carry out pilot production, verify the optimization effect, and fine-tune the process parameters according to the pilot results, and finally determine the optimized process plan for coloring artificial fur fabrics.

[0239] The process of coloring the artificial fur fabric using cationic dyes includes the following steps: preparation of grey cloth, water squeezing treatment, preparation of cationic dyes, preparation of high-temperature disperse dyes, integrated dyeing of cationic high-temperature disperse dyes, water-washing softener treatment, dehydration, drying, sizing and shaping of the base fabric, brushing, ironing, hot blowing, shaping, and ironing finishing.

[0240] The specific implementation methods of the above steps are described in detail below:

[0241] The specific implementation method of step S10 is to lay the foundation for subsequent process optimization through systematic data collection and sorting. First, collect the specification data of the grey cloth, including the raw material composition, pile height, width and gram weight of the weft knitted artificial fur. Secondly, record the water treatment parameters, such as water temperature, vehicle speed and pressure. Then, sort out the dye formula information, including the types and proportions of cationic dyes and high-temperature disperse dyes. Next, collect dyeing process parameters, such as dyeing temperature, time, bath ratio, etc. After that, summarize the post-processing process parameters, including the specific parameters of the steps of washing softener treatment, drying, base fabric sizing and shaping, brushing, ironing, hot blowing and shaping. Finally, collect the final product quality data, including indicators such as color fastness, feel, pile fastness and color uniformity. The collection of these data adopts a method combining production records, experimental tests and quality inspections to ensure the comprehensiveness and accuracy of the data. During the data collection process, standardized testing methods are used, such as ISO105 standard for color fastness, AATCC evaluation method for hand feel, friction test for fuzz fastness, and spectrophotometer for color uniformity. By establishing a complete database, reliable data support is provided for subsequent model establishment and optimization.

[0242] The specific implementation of step S20 is to establish an empirical equation group of coloring effect considering each process parameter. The purpose of this step is to digitize the complex dyeing process for subsequent optimization analysis. First, based on the dye diffusion theory, a dye diffusion equation is established to consider the concentration distribution of the dye in the fiber over time and space. The equation adopts the form of a partial differential equation, including diffusion coefficient, dye concentration gradient and dye adsorption and desorption terms. Secondly, a dye fixation equation is constructed to describe the chemical binding process between the dye and the fiber, considering the influence of temperature, pH value and auxiliaries. This equation usually adopts an exponential form and adds a periodic fluctuation term to simulate the fluctuation in actual production. Then, a fiber water absorption equation is established to describe the water absorption behavior of the fiber during the dyeing process, considering the influence of capillary action and fiber structure. Next, a fluff setting equation is constructed to describe the setting effect of the fluff during the heat treatment process, considering the influence of factors such as temperature, pressure and humidity. Afterwards, a color saturation equation is established to associate dye concentration, fiber characteristics and color space parameters to describe the color saturation of the final product. Finally, a process step correlation equation is constructed to describe the mutual influence and cumulative effect between each process step. The establishment of these equations adopts a method that combines theoretical analysis with experimental data fitting, and uses statistical methods such as least squares method and nonlinear regression to determine the coefficients and parameters in the equations. The establishment of the equation system provides a theoretical basis and initial framework for the subsequent machine learning model.

[0243] The specific implementation method of step S30 is to use a machine learning algorithm to train the established equation group using the collected process parameter data to obtain a preliminary process parameter-effect relationship model. First, the collected data is preprocessed, including data cleaning, normalization and standardization. Data cleaning includes removing outliers and processing missing values. The box plot method can be used to identify outliers and fill missing values ​​with the mean or median. Normalization uses the minimum-maximum scaling method to scale all features to the [0,1] interval. Standardization uses the Z-score method to make the data mean 0 and the standard deviation 1. Secondly, select a suitable machine learning algorithm. Considering the complexity and nonlinear characteristics of the dyeing process, algorithms such as support vector machine regression (SVR), random forest regression and gradient boosting tree (such as XGBoost) can be selected. SVR is suitable for processing high-dimensional feature space, random forest can handle the interaction between features, and XGBoost performs well in processing nonlinear relationships and feature importance ranking. Then, the data set is divided into a training set and a validation set, usually with a ratio of 80%:20%. Next, the model hyperparameters were tuned using grid search and cross-validation methods. For SVR, the kernel function type, regularization parameter C, and epsilon value were mainly tuned; for random forest, the number of trees, maximum depth, and minimum number of leaf node samples were tuned; for XGBoost, the learning rate, maximum depth, and subsampling ratio were tuned. During the training process, the root mean square error (RMSE) and determination coefficient (R²) were used as evaluation indicators. Finally, an explanatory analysis was performed on the trained model, using methods such as feature importance analysis and partial dependence diagrams to deeply understand the degree and manner of the influence of each process parameter on the dyeing effect. The purpose of this step is to combine the empirical equation with actual data to build a model that can accurately predict the relationship between process parameters and dyeing effects, and provide decision support for subsequent process optimization.

[0244] The specific implementation of step S40 is to generate a large number of process parameter combination data sets in a traversal manner according to the preset process parameter range and step length, and use the preliminary process parameter-effect relationship model for calculation to obtain a large number of candidate process parameter combinations and their theoretical effect data. First, determine the reasonable range and step length of each process parameter. For example, the dyeing temperature range can be set to 80°C to 120°C, with a step length of 5°C; the dyeing time range can be set to 30 minutes to 90 minutes, with a step length of 5 minutes; the dye concentration range can be set to 1g / L to 5g / L, with a step length of 0.5g / L. The setting of these ranges and step lengths needs to consider the feasibility and equipment limitations of actual production. Secondly, all possible parameter combinations are generated using the Cartesian product algorithm. Considering that the number of parameter combinations may be very large, parallel computing technology can be used to improve computing efficiency, such as using the ApacheSpark distributed computing framework. Then, for each set of parameter combinations, the machine learning model trained in step S30 is used for prediction to obtain theoretical effect data. The effect data here include the predicted values ​​of indicators such as color fastness, feel, velvet firmness and color uniformity. Next, the prediction results are screened to remove results that are obviously unreasonable or beyond the acceptable range. Some thresholds can be set, such as directly eliminating combinations with color fastness prediction values ​​lower than level 3 (with level 5 as the full score). After that, the remaining candidate process parameter combinations are preliminarily sorted, and the weighted average method can be used to assign different weights to each indicator according to its importance and calculate the comprehensive score. Finally, the top-N (such as top-1000) candidate process parameter combinations and their theoretical effect data are saved to prepare for subsequent cluster analysis. The purpose of this step is to quickly and comprehensively explore the process parameter space through computer simulation, discover potential optimization directions, and greatly reduce the number and cost of actual experiments.

[0245] The specific implementation method of step S50 is to perform cluster analysis on a large number of candidate process parameter combinations according to the theoretical effect data to obtain multiple cluster centers. First, the theoretical effect data is standardized to eliminate the dimensional differences between different indicators. The Z-score standardization method can be used to make the mean of each indicator 0 and the standard deviation 1. Secondly, select an appropriate clustering algorithm. Considering the high-dimensional characteristics of the data and possible non-spherical distribution, a hierarchical clustering algorithm or a density clustering algorithm can be selected. Specifically, the DBSCAN (density-based spatial clustering application noise) algorithm can be considered, which can find clusters of any shape and can automatically detect and remove noise points. For the DBSCAN algorithm, two key parameters need to be set: ε (neighborhood radius) and MinPts (minimum number of points). The appropriate ε value can be determined by the K-distance graph method, and MinPts is usually set to twice the data dimension. Then, run the clustering algorithm to obtain preliminary clustering results. Next, evaluate and optimize the clustering results. The quality of clustering can be evaluated using the Silhouette Coefficient, which measures the difference between the similarity of a sample to its own cluster and the similarity of other clusters. The clustering results are iteratively optimized by adjusting clustering parameters such as ε and MinPts, and trying different clustering algorithms (such as K-means++ or OPTICS algorithm). After that, feature analysis is performed on each cluster, the cluster center is calculated, and the characteristics of each cluster are explained. The principal component analysis (PCA) method can be used to reduce the dimension of high-dimensional data to 2D or 3D space for visualization to help understand the clustering structure. Finally, representative process parameter combinations are selected based on the clustering results. The actual parameter combination closest to the cluster center can be selected, or the parameter value of the cluster center can be directly used (appropriate rounding processing is required). Generally, it is appropriate to select 5 to 10 cluster centers as the process set to be tested. The purpose of this step is to classify a large number of candidate process parameter combinations, find representative parameter combinations, and provide a scientific experimental design plan for subsequent small-scale production tests.

[0246] The specific implementation method of step S60 is to select the process parameter combination corresponding to each cluster center as the process set to be tested, conduct a small-scale production test, and collect actual effect data. First, according to the cluster center obtained in step S50, a small-scale production test plan is designed. Considering the limitations of production resources and time, 5 to 10 representative process parameter combinations are usually selected for testing. Secondly, prepare experimental materials and equipment. Ensure the batch consistency of raw materials (such as grey cloth, dyes, auxiliaries, etc.) and the stability of equipment (such as dyeing machines, dryers, setting machines, etc.). Then, according to the designed process parameter combination, small-scale production tests are carried out in sequence. Each group of tests requires strict control of process parameters, such as dyeing temperature, time, dye concentration, etc., to ensure consistency with the design plan. At the same time, record real-time data in the production process, such as temperature curves, pH changes, etc. Next, a comprehensive quality test is carried out on the produced samples. The test items include color fastness (washing, friction, sun exposure, etc.), feel (softness, stiffness, etc.), fuzz fastness (anti-pilling) and color uniformity. The color fastness test adopts the ISO 105 standard method, the hand feel evaluation can be objectively measured using the KES-F system, the fuzz fastness can be tested by the Martindale method, and the color uniformity is measured at multiple points using a spectrophotometer. After that, the test data is statistically analyzed, the mean and standard deviation of each index are calculated, and the reliability and stability of the data are evaluated. If the degree of dispersion of a set of data is too large (such as the coefficient of variation exceeds 10%), it is necessary to consider repeating the experiment. Finally, the actual effect data is compared and analyzed with the theoretical prediction data, the prediction error is calculated, and the accuracy of the model is evaluated. Indicators such as root mean square error (RMSE) and mean absolute percentage error (MAPE) can be used to quantify the prediction performance of the model. The purpose of this step is to verify the reliability of the theoretical model through actual production, and at the same time collect real effect data to provide a basis for subsequent model optimization and process parameter adjustment.

[0247] The specific implementation method of step S70 is to compare the actual effect data with the theoretical effect data, optimize the process parameter-effect relationship model, and improve the accuracy and reliability of the model. First, the actual effect data collected in step S60 is paired with the theoretical prediction data in step S40. For each process parameter combination, the error between the actual value and the predicted value is calculated, including the absolute error and the relative error. Secondly, the error distribution characteristics are analyzed, such as whether there is a systematic deviation, whether the error changes with the parameter change, etc. Visualization methods such as box plots and scatter plots can be used to intuitively display the error distribution. Then, according to the error analysis results, the original model is adjusted and optimized. If there is a systematic deviation, you can consider adding a bias term or introducing a nonlinear transformation; if the error changes with the parameter change, you can consider introducing an interaction term or using a more complex nonlinear model. Specific optimization methods may include: 1) Feature engineering: create new features, such as square terms of parameters, cross terms, etc., to capture more complex nonlinear relationships. 2) Model structure adjustment: If a neural network model is used, you can try to adjust the number of network layers and the number of neurons; if it is a tree model, you can adjust the depth of the tree, the minimum number of samples of leaf nodes, etc. 3) Ensemble learning: Combine multiple basic models, such as using ensemble methods such as random forest or gradient boosting tree, to improve the generalization ability of the model. 4) Regularization: Introduce L1 or L2 regularization terms to prevent model overfitting. 5) Transfer learning: Use a model pre-trained on a large amount of historical data to adapt to the current specific task through fine-tuning. Next, use the cross-validation method to retrain and evaluate the optimized model. K-fold cross-validation (usually k=5 or 10) can be used to evaluate the performance stability of the model on different data subsets. Evaluation indicators include root mean square error (RMSE), mean absolute error (MAE), determination coefficient (R²), etc. After that, compare the performance of the model before and after optimization. If the optimized model performs significantly better than the original model on the validation set (such as RMSE is reduced by more than 10%), the new model is adopted; otherwise, the cause needs to be further analyzed, and more experimental data may need to be collected or other optimization strategies may need to be tried. Finally, the optimized model is used to re-predict the effect data of a large number of process parameter combinations generated in step S40, and the candidate process parameter set is updated. The purpose of this step is to continuously iterate and improve the model through actual data feedback, so that it is closer to the actual production situation and improves the accuracy and reliability of the prediction.

[0248] The specific implementation method of step S80 is to use multiple groups of process parameters with the best actual effect in the process to be tested as the initial population, use the comprehensive effect score as the fitness function, set the crossover rate and mutation rate, and use a genetic algorithm to obtain the optimal process parameter combination through multi-generation iterative optimization. First, select 3 to 5 groups of process parameters with the best actual effect as the initial population from the experimental results of step S60. The effect evaluation adopts a comprehensive scoring method to perform weighted averaging on indicators such as color fastness, feel, fuzz fastness and color uniformity. The weights can be determined based on product demand and market feedback, for example, the color fastness weight is 0.3, the feel weight is 0.3, the fuzz fastness weight is 0.2, and the color uniformity weight is 0.2. Secondly, design the fitness function. The fitness function should be able to reflect the comprehensive performance of the process parameter combination, which can be defined as: ,in is the weight of each indicator, is the normalized score of each indicator. The score normalization can use the minimum-maximum scaling method to map each indicator to the [0,1] interval. Then, set the parameters of the genetic algorithm. The population size can be set to 50 to 100, the crossover rate is usually set between 0.6 and 0.8, and the mutation rate is set between 0.01 and 0.1. The selection operation uses the tournament selection method, the crossover operation uses the single-point crossover, and the mutation operation uses the Gaussian mutation. The number of iterations can be set to 100 to 500 generations, or stop when the optimal solution has not improved for 30 consecutive generations. Next, implement the main steps of the genetic algorithm: 1) Initialize the population: In addition to the optimal combination selected from the experimental results, randomly generate other individuals to ensure the diversity of the population. 2) Evaluate fitness: Use the model optimized in step S70 to predict the effect of each individual and calculate the fitness value. 3) Selection operation: Use the tournament selection method to select excellent individuals. 4) Crossover operation: Perform single-point crossover on the selected individuals to generate new offspring. 5) Mutation operation: Perform Gaussian mutation on some individuals to introduce randomness. 6) Update the population: add the newly generated offspring to the population and eliminate individuals with lower fitness. 7) Repeat steps 2) to 6) until the termination condition is met. After that, conduct a feasibility analysis on the optimal process parameter combination finally obtained. Check whether the parameters are within the range allowed by the equipment and whether they are in line with production reality. If there are infeasible parameters, they need to be adjusted or the algorithm needs to be rerun. Finally, output the optimal process parameter combination and its predicted comprehensive effect score. The purpose of this step is to search for the optimal solution among a large number of possible process parameter combinations by simulating the evolutionary process and find the theoretically best process parameter combination.

[0249] The specific implementation method of step S90 is to use the optimal process parameter combination to carry out pilot production, verify the optimization effect, and fine-tune the process parameters according to the pilot results, and finally determine the optimized process plan for coloring artificial fur fabrics. First, according to the optimal process parameter combination obtained in step S80, a detailed pilot production plan is formulated. The plan should include raw material preparation, equipment debugging, process flow, quality control points and other contents. Considering the changes in production scale, some parameters may need to be appropriately adjusted, such as dye bath ratio, heating rate, etc. Secondly, preparatory work before pilot production is carried out. Including raw material inspection (such as fiber quality, dye purity, etc.), equipment calibration (such as temperature sensor, pressure gauge, etc.), personnel training, etc. Ensure that all conditions are consistent with the small test stage to facilitate the comparability of the results. Then, pilot production is carried out according to the formulated plan. During the production process, various process parameters, such as dyeing temperature, time, pH value, etc., are strictly controlled, and changes in key process parameters are monitored and recorded in real time. At the same time, sampling is carried out at key process nodes to detect intermediate products and promptly discover and solve possible problems. Then, a comprehensive quality inspection is carried out on the products produced in the pilot production. The test items include color fastness (washing, friction, sun exposure, etc.), hand feel (softness, stiffness, etc.), velvet fastness (anti-pilling), color uniformity, etc. The test method should be consistent with the pilot stage to ensure the comparability of the data. At the same time, some tests simulating actual use are added, such as performance changes after repeated washing, wearing comfort evaluation, etc. After that, the pilot results are compared and analyzed with the pilot results. The differences in various indicators are calculated and the reasons for the differences are analyzed. Possible reasons include: differences in heat transfer and mass transfer caused by changes in production scale; differences in equipment accuracy; differences in raw material batches, etc. According to the analysis results, the process parameters are fine-tuned. The principle of fine-tuning is to maximize production efficiency and reduce energy consumption and costs while ensuring product quality. The orthogonal experimental method can be used to design fine-tuning experiments to quickly find the best parameter combination. Finally, the final optimization process plan is determined by comprehensively considering factors such as product quality, production efficiency, and cost. The plan should include detailed process parameters, operating procedures, quality control standards, etc. At the same time, the allowable fluctuation range of process parameters is formulated to provide guidance for subsequent large-scale production. The purpose of this step is to apply the theoretical optimization results to actual production, verify its feasibility and effectiveness, and further improve the process plan through practical tests, and finally form a reliable and popularizable artificial fur fabric coloring optimization process.

[0250] In order to better understand the present invention, a specific embodiment 1 of the present invention is provided below. The steps of this embodiment 1 are described in detail as follows: Step S10: Collecting various parameter data of the artificial fur fabric coloring process

[0251] In this step, a large amount of process parameter data needs to be collected and organized. This data can be expressed in terms of matrix express:

[0252] ,

[0253] in, represents the number of samples collected, Indicates the number of parameters. Each line Represents a complete set of process parameters.

[0254] For quality data, we can use the vector express:

[0255] ,

[0256] in, Indicates The quality score corresponding to the set of process parameters.

[0257] To assess the reliability of the data, the coefficient of variation for each parameter can be calculated :

[0258] ,

[0259] in, and Respectively The standard deviation and mean of the parameters. , further inspection and verification of this parameter is required.

[0260] Step S20: Establishing a coloring effect empirical equation group considering various process parameters

[0261] In this step, a series of equations need to be established to describe the relationship between process parameters and dyeing effect. These equations have been given in the previous description and will not be repeated here.

[0262] Step S30: Fitting the coloring effect empirical equations using the collected parameter data

[0263] In this step, a machine learning algorithm is needed to fit the process parameter-effect relationship model. Taking support vector machine regression (SVR) as an example, its objective function can be expressed as:

[0264] ,

[0265] in:

[0266] ,

[0267] ,

[0268] .

[0269] in, is the weight vector, is the bias term, and is the slack variable, is the penalty parameter, is the allowable error range, It is a function that maps the input space to a high-dimensional feature space.

[0270] For random forest regression, Gini impurity can be used to measure the purity of a node:

[0271] ,

[0272] in, Represents a node, Representation Node Medium Category The sample proportion.

[0273] For gradient boosting trees (such as XGBoost), the objective function can be expressed as:

[0274] ,

[0275] in, is the loss function, is the regularization term, is the number of leaf nodes, and is the regularization parameter.

[0276] During model training, the root mean square error (RMSE) and the coefficient of determination (R²) are used as evaluation metrics:

[0277] ,

[0278] ,

[0279] in, is the actual value, is the predicted value, is the average of the actual values.

[0280] Step S40: Generate a large number of process parameter combination data sets by traversing according to the preset process parameter range and step size

[0281] In this step, all possible parameter combinations need to be generated. Assume that parameters, each with possible values, the total number of combinations for:

[0282] .

[0283] For each set of parameter combinations, use the trained model to make predictions. Taking SVR as an example, the prediction function can be expressed as:

[0284] ,

[0285] in, and is the Lagrange multiplier, is the kernel function.

[0286] For the screening of prediction results, a threshold can be set , only keep the combinations that meet the following conditions:

[0287] .

[0288] For the remaining candidate process parameter combinations, the weighted average method is used to calculate the comprehensive score:

[0289] ,

[0290] in, It is The weight of the indicator, It is The predicted value of an indicator.

[0291] Step S50: Perform cluster analysis on a large number of candidate process parameter combinations based on theoretical effect data

[0292] In this step, the theoretical effect data is first standardized. The Z-score standardization method is used:

[0293] ,

[0294] in, It is The sample eigenvalues, and They are The mean and standard deviation of the features.

[0295] For the DBSCAN algorithm, two key parameters need to be defined: (neighborhood radius) and (Minimum number of points). For any point ,definition:

[0296] ,

[0297] in, is a dataset, Yes and The distance between.

[0298] if ,but It is called the core point.

[0299] The main steps of the DBSCAN algorithm are as follows:

[0300] For each unvisited point ,calculate ;

[0301] if is a core point, a new cluster is created and All points in are added to the cluster;

[0302] For each point newly added to the cluster, if it is also a core point, repeat step 2;

[0303] If all points have been visited, the algorithm terminates.

[0304] To evaluate the quality of clustering, the Silhouette Coefficient is used:

[0305] ,

[0306] in, It is a sample The average distance to other samples in the same cluster, It is a sample The average distance to the nearest samples in other clusters.

[0307] Step S60: Select the process parameter combination corresponding to each cluster center as the process set to be tested and conduct a small-scale production test

[0308] In this step, a small-scale production test needs to be designed. Assuming that Cluster centers, each center performs The total number of experiments is .

[0309] For each set of experiments, real-time data, such as temperature curves, need to be recorded. and pH changes :

[0310] ,

[0311] ,

[0312] in, and are the initial temperature and pH value, and is the amount of change, and is the rate constant of change.

[0313] For quality test data, the average value can be calculated and standard deviation :

[0314] ,

[0315] .

[0316] If the coefficient of variation of a set of data , it is necessary to consider repeating the experiment.

[0317] Step S70: Compare the actual effect data with the theoretical effect data and optimize the process parameter-effect relationship model

[0318] In this step, the error between the actual value and the predicted value is first calculated:

[0319] Absolute error: ,

[0320] Relative error: .

[0321] For error analysis, a box plot can be used to visualize the error distribution. The five key values ​​of a box plot are:

[0322] Minimum: ,

[0323] Lower quartile: ,

[0324] Median: ,

[0325] Upper quartile: ,

[0326] Maximum value: .

[0327] For model optimization, the following methods can be considered:

[0328] Feature engineering: creating new features, such as , wait;

[0329] Model structure adjustment: For neural networks, the number of hidden layers can be adjusted and the number of neurons in each layer ;

[0330] Ensemble learning: Use methods such as random forest or gradient boosting tree. The prediction function of random forest can be expressed as:

[0331] ,

[0332] in, is the number of trees, It is The prediction results of each tree;

[0333] Regularization: Introduce L1 or L2 regularization terms. Taking L2 regularization as an example, the objective function becomes:

[0334] ,

[0335] in, is the regularization parameter;

[0336] Transfer learning: Use a model pre-trained on a large amount of historical data to adapt it to the current task through fine-tuning. The fine-tuning process can be expressed as:

[0337] ,

[0338] in, are the parameters of the pre-trained model, is the learning rate.

[0339] During model evaluation, k-fold cross validation is used (usually k=5 or 10). For each fold, RMSE and R² are calculated:

[0340] ,

[0341] .

[0342] The final evaluation metric is the average of k calculations:

[0343] ,

[0344] .

[0345] Step S80: Using genetic algorithm to obtain the optimal process parameter combination through multiple generations of iterative optimization

[0346] In this step, a genetic algorithm is used to optimize the process parameter combination. Genetic algorithm is an optimization algorithm that simulates the biological evolution process and mainly includes the following key steps:

[0347] 1. Initialize the population:

[0348] First, select multiple sets of process parameters with the best actual effect from the process set to be tested as the initial population. Assume that individuals, each of which contains process parameters, the initial population can be expressed as:

[0349] ,

[0350] in, Indicates The initial parameter combination of each individual.

[0351] 2. Calculate fitness:

[0352] The fitness function is the key to evaluating the quality of an individual. In this case, the comprehensive effect score is used as the fitness function:

[0353] ,

[0354] in, They are the weights of color fastness, hand feel, nap firmness and color uniformity. are the normalized scores of these four indicators respectively.

[0355] The normalized score can be calculated using min-max scaling:

[0356] ,

[0357] in, It is The individual in The raw scores on the indicators are and They are The minimum and maximum values ​​of the indicators.

[0358] 3. Select an action:

[0359] Use tournament selection. Randomly select from the population individuals, and select the one with the highest fitness. This process is repeated times, to select individuals as parents.

[0360] The probability of tournament selection can be expressed as:

[0361] ,

[0362] in, Is an individual The fitness value of is the number of individuals participating in the tournament.

[0363] 4. Crossover operation:

[0364] Use single-point crossover. For the two selected parent individuals and , randomly select an intersection point ( ), and then exchange the genes of the two parents after this point to generate two offspring individuals and :

[0365] ,

[0366] .

[0367] The probability of a crossover operation is , usually set between 0.6 and 0.8.

[0368] 5. Mutation operation:

[0369] Gaussian mutation is used. For the gene selected for mutation , mutate according to the following formula:

[0370] ,

[0371] in, The mean is 0 and the standard deviation is The normal distribution of . It can be set to 1 / 10 to 1 / 5 of the parameter value range.

[0372] The probability of mutation operation is , usually set between 0.01 and 0.1.

[0373] 6. Update population:

[0374] Add the newly generated offspring individuals to the population and select the top Individuals serve as the new generation population.

[0375] 7. Termination conditions:

[0376] Set the maximum number of iterations , or when continuous Stop when there is no improvement in the optimal solution after 30 generations (for example).

[0377] The iterative process can be expressed as:

[0378] ,

[0379] in, Indicates Generation population, , and Represent selection, crossover and mutation operations respectively.

[0380] In each iteration, record the current optimal solution And the corresponding fitness value :

[0381] ,

[0382] .

[0383] In order to improve the convergence speed and global search capability of the algorithm, the following strategies can be adopted:

[0384] a) Adaptive crossover and mutation probability:

[0385] As the iterations proceed, the crossover probability and mutation probability are gradually reduced:

[0386] ,

[0387] ,

[0388] in, and are the initial crossover probability and mutation probability, respectively. is the current iteration number.

[0389] b) Elite Retention Strategy:

[0390] In each generation, the one with the highest fitness is retained Individuals ( ) No crossover and mutation operations are performed and the next generation is directly entered.

[0391] c) Restart mechanism:

[0392] If continuous If the optimal solution of the previous generation has not been improved, the population is reinitialized but the current optimal solution is retained.

[0393] d) Local search:

[0394] In the genetic algorithm, Every 10 generations (for example), perform a local search for the current optimal solution. You can use the simulated annealing algorithm for local search:

[0395] ,

[0396] in, is the search step length, is a standard normal distribution.

[0397] The acceptance probability is:

[0398] ,

[0399] in, is the temperature parameter of simulated annealing, which decreases gradually with the iterations:

[0400] ,

[0401] in, is the initial temperature, is the temperature reduction coefficient (usually between 0.95 and 0.99).

[0402] In the whole process of genetic algorithm, the following points need to be noted:

[0403] 1) Parameter encoding: For continuous parameters, real number encoding can be used directly. For discrete parameters, integer encoding or binary encoding can be used.

[0404] 2) Constraint processing: After the crossover and mutation operations, it is necessary to check whether the parameters exceed the allowed range. If so, the following strategies can be used:

[0405] Boundary Absorption: Set out-of-range parameter values ​​to boundary values.

[0406] Reflection: Reflect the out-of-range portion back into the feasible area.

[0407] Regenerate: Regenerate a random value within the allowed range.

[0408] 3) Design of fitness function: In addition to considering color fastness, hand feel, nap firmness and color uniformity, other factors can also be introduced, such as production cost, energy consumption, etc. The weighted sum method or the analytic hierarchy process (AHP) can be used to determine the weight of each indicator.

[0409] 4) Diversity maintenance:

[0410] In order to prevent the population from reaching the local optimal solution too early, the following strategies can be adopted:

[0411] Island model: The population is divided into multiple sub-populations, and individuals migrate regularly.

[0412] Crowding penalty: Introducing a crowding penalty term into the fitness calculation to encourage individuals to disperse in the solution space.

[0413] 5) Parallelization: To improve the efficiency of the algorithm, parallel computing technology can be used. For example, the fitness of multiple individuals can be evaluated at the same time, or multiple independent genetic algorithm instances can be run in parallel and the optimal result can be selected.

[0414] 6) Parameter sensitivity analysis: For the optimal process parameter combination obtained, parameter sensitivity analysis is performed to understand the influence of each parameter on the final result:

[0415] ,

[0416] in, It is The sensitivity coefficients of the parameters.

[0417] 7) Result visualization: Use parallel coordinates or radar charts to visualize the optimal process parameter combination and intuitively display the values ​​of each parameter.

[0418] 8) Convergence analysis: Draw the curve of the optimal fitness value and the average fitness value during the iteration process to analyze the convergence of the algorithm:

[0419] ,

[0420] .

[0421] Through the above steps, the theoretically optimal combination of process parameters can be obtained. However, it should be noted that the results obtained by the genetic algorithm may have a certain degree of randomness. Therefore, it is recommended to run the algorithm multiple times and compare and comprehensively analyze the multiple sets of results obtained.

[0422] Step S90: Use the optimal process parameter combination to conduct pilot production, verify the optimization effect, and fine-tune the process parameters according to the pilot results to ultimately determine the optimized process solution for coloring the artificial fur fabric.

[0423] Specifically, the principle of the present invention is to transform the traditional experience-driven optimization into an intelligent optimization that combines data-driven and mechanism models. This method can effectively solve the technical problems existing in the coloring process of artificial fur fabrics, and its principle can be explained from the following aspects:

[0424] 1. Combination of mechanism model and data drive: The present invention first establishes a set of empirical equations for coloring effects that take into account various process parameters. These equations are based on the physical and chemical principles of the dyeing process, such as dye diffusion theory, adsorption kinetics, etc. The introduction of this mechanism model provides theoretical guidance for subsequent data-driven optimization, making the optimization process no longer a pure "black box" operation. At the same time, training actual data through machine learning algorithms can make up for the shortcomings of theoretical models and capture various non-ideal factors in actual production. This combination not only ensures the theoretical basis of the model, but also improves its adaptability to actual conditions.

[0425] 2. Effective exploration of high-dimensional parameter space: The artificial fur fabric coloring process involves multiple steps, each of which contains multiple parameters, forming a high-dimensional parameter space. The present invention uses a traversal method to generate a large number of process parameter combinations, and uses a trained model for rapid prediction. This method can explore a large number of parameter combinations in a short time. Through cluster analysis, similar parameter combinations are classified, which not only ensures the comprehensiveness of the search, but also avoids redundant experiments, greatly improving the optimization efficiency.

[0426] 3. Balance of multi-objective optimization: The quality of artificial fur fabrics involves multiple indicators, such as color fastness, hand feel, nap firmness and color uniformity. There may be contradictory relationships between these indicators, and the optimization of a single indicator may lead to the deterioration of other indicators. The present invention converts multiple objectives into a single optimization objective by designing a comprehensive effect score as a fitness function, thereby achieving a balanced optimization of multiple objectives. This method meets actual production needs and can obtain more practical optimization results.

[0427] 4. Closed-loop feedback and continuous optimization: The present invention collects actual effect data through small-scale production tests, compares it with the theoretical prediction results, and continuously optimizes and improves the model. This closed-loop feedback mechanism enables the model to continuously learn and adapt to actual production conditions. As data accumulates, the prediction accuracy of the model will continue to improve, and the optimization results will become closer and closer to the actual optimal solution. This continuous optimization mechanism makes the method of the present invention highly adaptable and scalable.

[0428] 5. Application of intelligent algorithms: The present invention uses a variety of intelligent algorithms, such as support vector machine regression, random forest regression, gradient boosting tree and other machine learning algorithms, as well as evolutionary computing methods such as genetic algorithms. These algorithms have significant advantages in dealing with complex nonlinear problems and high-dimensional parameter optimization. In particular, the application of genetic algorithms, by simulating the biological evolution process, can quickly converge to the global optimal solution in a complex parameter space, avoiding the risk of falling into the local optimum.

[0429] 6. Knowledge accumulation and experience inheritance: The method of the present invention can not only obtain optimized process parameters, but also reveal the influence mechanism of each parameter on the dyeing effect through model interpretative analysis. This interpretability makes the optimization process no longer a "black box", but can generate understandable and verifiable knowledge. This knowledge can be used as experience accumulation to guide future production and research and development, and realize the inheritance and accumulation of knowledge.

[0430] In summary, the technical solution of the present invention combines the mechanism model with the data-driven method, utilizes the intelligent algorithm for efficient optimization, and achieves continuous improvement through closed-loop feedback, thus forming a process optimization system with rigorous logic and the combination of theory and practice.

[0431] In order to further better understand and implement the present invention, Example 2 of a specific application scenario of the present invention is provided below: This Example 2 takes the imitation mink artificial fur fabric produced by a textile factory as an example, and adopts the method of the present invention to optimize its coloring process.

[0432] 1. Data Collection (S10)

[0433] First, we collected 100 sets of historical data of the factory producing this type of product in the past year, including the following parameters:

[0434] 1.1 Grey fabric specifications: raw material of pile: polyester fiber; raw material of base fabric: polyester plus stretch yarn; organizational structure: weft knitted organization; pile height: 15-90mm; net width: 160-185cm; gram weight: 800-3000g / m²;

[0435] 1.2 Water treatment parameters: water temperature: 40-50℃; speed: 30-50 yards / minute; pressure: 1.5-2.5KG;

[0436] 1.3 Dye formula: Cationic dyes: SD golden yellow, red, blue (0.5-2% each); high temperature disperse dyes: disperse red, orange, yellow (0.3-1.5% each);

[0437] 1.4 Dyeing process parameters: bath ratio: 1:8-1:12; dyeing temperature: 30-40℃; heating rate: 1-2℃ / min; holding temperature: 120-130℃; holding time: 30-60min; cooling rate: 1-2℃ / min;

[0438] 1.5 Post-processing process parameters include the following:

[0439] Washing and softening agent treatment: water ratio 1:4, softener 2-5%, static agent 0.5-1.5%, speed 30-50 yards / minute, pressure 1.5-2.5KG, temperature 40-50℃;

[0440] Drying: speed 35-55 yards / minute, temperature 140-180℃;

[0441] Sizing and shaping of base fabric: slurry concentration 25-80%, temperature 170-200℃, vehicle speed 30-45m / min, wind speed 600-1450 rpm;

[0442] Brushing: speed 4-8 yards / min, tension 1020-1032, speed 1000-11000 rpm, brushing depth 1.0-2.4mm;

[0443] Ironing: speed 12-16m / min, temperature 120-200℃, speed 500-700 rpm;

[0444] Hot blowing: speed 0.5-1.0m / min, temperature 130-140℃;

[0445] Shaping: speed 40-50m / min, temperature 120-130℃, wind speed 180-220, width 1950-2000mm;

[0446] Ironing finishing: speed 12-16m / min, temperature 120-200℃, speed 500-700 rpm;

[0447] 1.6 Quality Data:

[0448] Color fastness: color fastness to washing, color fastness to rubbing, color fastness to light (grade 1-5);

[0449] Hand feel: softness, fluffiness (level 1-5);

[0450] Fluff firmness: 1-5 levels;

[0451] Color uniformity: 1-5 levels;

[0452] 2. Establishment of empirical equations (S20)

[0453] Based on the collected data, the following set of empirical equations was established:

[0454] 2.1 Dye diffusion equation:

[0455] ;

[0456] 2.2 Dye fixation equation:

[0457] ;

[0458] 2.3 Fiber water absorption equation:

[0459] ;

[0460] 2.4 Villus shaping equation:

[0461] ;

[0462] 2.5 Color saturation equation:

[0463] ;

[0464] 2.6 Process step correlation equation:

[0465] ;

[0466] 3. Model fitting (S30)

[0467] The empirical equations were fitted using the collected historical data to obtain a preliminary model. Taking the dye diffusion equation as an example, the following parameters were obtained through nonlinear regression: D = 5.2 × 10^-11 m² / s; k1 = 0.015 L / (g·s); k2 = 0.003 1 / s; Cmax = 25 g / L;

[0468] 4. Generate candidate process parameter combinations (S40)

[0469] According to the preset process parameter range and step length, 10,000 sets of candidate process parameter combinations were generated. Taking the water-squeezing process as an example, the parameter range and step length are as follows: water temperature: 40-50℃, step length 1℃; vehicle speed: 30-50 yards / minute, step length 1 yard / minute; pressure: 1.5-2.5KG, step length 0.1KG;

[0470] 5. Cluster analysis (S50)

[0471] K-means cluster analysis was performed on 10,000 sets of candidate process parameter combinations, K=10 was selected, and 10 cluster centers were obtained.

[0472] 6. Small-scale production trial (S60)

[0473] The process parameter combinations corresponding to 10 cluster centers were selected for small-scale production tests, with 3 tests per group and a total of 30 tests. Actual effect data were collected, including color fastness, hand feel, nap firmness and color uniformity.

[0474] 7. Model optimization (S70)

[0475] The actual effect data were compared with the theoretical effect data, and the preliminary model was optimized using the least squares method to obtain the process parameter-effect relationship optimization model.

[0476] 8. Genetic algorithm optimization (S80)

[0477] The three groups of process parameters with the best effect in the small-scale test were used as the initial population, and the following parameters were set: population size: 50; crossover rate: 0.8; mutation rate: 0.1; number of iterations: 100;

[0478] The fitness function is a comprehensive effect score, and the weights are as follows: color fastness: 40%; feel: 30%; fuzz fastness: 20%; color uniformity: 10%;

[0479] After 100 generations of iteration, the optimal process parameter combination is obtained: the specific description is as follows: the process steps include the following steps: grey cloth preparation, water pressing treatment, cationic dye preparation, high temperature disperse dye preparation, cationic high temperature disperse integrated dyeing, water washing softener treatment, dehydration, drying, base cloth sizing and shaping, brushing, ironing, hot blowing, shaping, ironing finishing, wherein: in the grey cloth preparation step, the grey cloth is weft knitted artificial fur, the raw material: the fluff is polyester fiber, the base cloth raw material is polyester plus stretch yarn, the tissue structure is weft knitted tissue, the fluff height is 15~90mm, the net width is 160~185cm, and the gram weight is 800~3000g / m; in the water-squeezing step, (water-squeezing, water temperature (40-50°C), vehicle speed (30-50 yards), pressure (2KG)); in the preparation step of the printing cationic dye, (cationic dye formula: SD golden yellow, red, blue); in the preparation step of the high-temperature disperse dye, (high-temperature disperse dye formula: dispersed material red, orange, yellow); in the cationic high-temperature dispersed~body dyeing step, (bath ratio, dyeing temperature, heating time, insulation time, insulation temperature, cooling time range); in the washing softener treatment step, (water ratio 1:4, softener ratio 2%-5%, electrostatic agent ratio 1%, vehicle speed 30-50 yards, pressure 2KG, temperature range 40℃-50℃); in the drying step, (speed 35-55 yards, temperature range 160℃ / 150℃); in the base fabric sizing and shaping step, (slurry formula 25%-80%, temperature 170-200℃, speed 30m / min-45m / min, wind speed 600rpm-1450rpm, up and down wind range: 80% product downwind, 20% product upwind); in the brushing step, (speed 6 yards, tension of each roller 1032-1020, speed 1000-11000, depth of each roller into 2.2 1.2); in the ironing step, (process parameter range of each online equipment) speed (14) temperature (188-120℃) speed (600); in the hot blowing step, (shrinking process parameter range) speed (0.7)) temperature (136℃); in the shaping step, (stretching process parameter range) speed (43) temperature (125℃) wind force (200) width (1970); in the ironing step, (long-line ironing process parameter range) speed (14) temperature (two types of 188-120 / 120-80℃) speed (600).

[0480] 9. Pilot production and fine-tuning (S90)

[0481] The optimal process parameter combination was used for three pilot production runs, with 500 meters of fabric produced each time. The process parameters were fine-tuned based on the pilot results, and the final optimized process solution was determined as shown in Table 2 below:

[0482] Table 2 Final process plan

[0483]

[0484]

[0485] After optimization, product quality is significantly improved. Figure 2 As shown:

[0486] 1. Color fastness: Color fastness to washing is improved from level 3-4 to level 4-5;

[0487] The color fastness to rubbing is increased from level 3 to level 4;

[0488] Color fastness to light increased from level 3 to level 4;

[0489] 2. Hand feel: Softness increased from level 3 to level 4-5; Fluffiness increased from level 3-4 to level 4-5;

[0490] 3. Fluff firmness: increased from level 3 to level 4-5;

[0491] 4. Color uniformity: improved from level 3-4 to level 4-5.

[0492] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A process optimization method for coloring artificial fur fabric using cationic, characterized in that: The following steps are involved: S10, collecting various parameter data of the artificial fur fabric coloring process, including grey fabric specifications, squeezing treatment parameters, dye formula, dyeing process parameters, post-treatment process parameters and final product quality data; S20, establishing a group of empirical equations for coloring effects taking into account various process parameters; S30, fitting the coloring effect empirical equation group using the collected parameter data to obtain a preliminary process parameter-effect relationship model; S40, generating a large number of process parameter combination data sets in a traversal manner according to a preset process parameter range and step size, and calculating using the process parameter-effect relationship preliminary model to obtain a large number of candidate process parameter combinations and their theoretical effect data; S50, performing cluster analysis on a large number of candidate process parameter combinations according to theoretical effect data to obtain multiple cluster centers; S60, selecting a process parameter combination corresponding to each cluster center as a process set to be tested, conducting a small-scale production test, and collecting actual effect data; S70, comparing the actual effect data with the theoretical effect data, optimizing the preliminary model of the process parameter-effect relationship, and obtaining a process parameter-effect relationship optimization model; S80, taking multiple groups of process parameters with the best actual effects in the process to be tested as the initial population, setting the crossover rate and mutation rate, and using a genetic algorithm to obtain the optimal process parameter combination through multiple generations of iterative optimization; S90. Use the optimal process parameter combination to carry out pilot production, verify the optimization effect, and fine-tune the process parameters according to the pilot results, and finally determine the optimized process plan for coloring artificial fur fabrics.

2. The process optimization method for coloring artificial fur fabric using cations according to claim 1 is characterized in that: The quality data include color fastness, hand feel, pile fastness and color uniformity.

3. The process optimization method for coloring artificial fur fabric using cationic materials according to claim 2, characterized in that: The coloring effect empirical equation group includes a dye diffusion equation, a dye fixation equation, a fiber water absorption equation, a fluff setting equation, a color saturation equation and a correlation equation between various process steps.

4. The process optimization method for coloring artificial fur fabric using cations according to claim 3 is characterized in that: The method of fitting the coloring effect empirical equation group using the collected parameter data is the least squares fitting.

5. The process optimization method for coloring artificial fur fabric using cations according to claim 4 is characterized in that: The preset process parameter range and step size, wherein the step size is 0.01~0.1 of the corresponding process parameter range.

6. The process optimization method for coloring artificial fur fabric using cations according to claim 5 is characterized in that: K-means clustering was used for cluster analysis.

7. The process optimization method for coloring artificial fur fabric using cations according to claim 6 is characterized in that: The genetic algorithm uses the comprehensive effect score obtained by the process parameter-effect relationship optimization model as a fitness function.

8. The process optimization method for coloring artificial fur fabric using cations according to claim 7 is characterized in that: The comprehensive effect score is a weighted average of color fastness, feel, fuzz fastness and color uniformity.

9. The process optimization method for coloring artificial fur fabric using cations according to claim 8, characterized in that: The iteration stopping condition of the genetic algorithm includes reaching the maximum number of iteration steps of 2000 times.

10. The process optimization method for coloring artificial fur fabric using cations according to claim 8, characterized in that: The iteration stopping condition of the genetic algorithm is that the difference between the fitness function values ​​of two consecutive iteration results is less than 0.00375.

Images