Blended yarn process intelligent design system combining digital twinning and multi-objective optimization

Abstract

The invention provides a blended yarn process intelligent design system combining digital twinning and multi-objective optimization, relates to the field of textile process optimization, and improves the efficiency of blended yarn process design and the product quality stability. The method comprises the following steps: firstly, collecting real-time data, establishing a digital twinning model, simulating multiple groups of process parameters in a virtual environment, predicting yarn quality data, and obtaining a feasible parameter space; thirdly, calculating the influence weight of the process parameters on the yarn quality data in the feasible parameter space, establishing a multi-objective optimization function to generate a Pareto optimal solution set, and further obtaining the optimal process parameters; and finally, calculating and adjusting the priority according to the parameter deviation, sequentially adjusting the process parameters, comparing the predicted yarn quality data with the actual yarn quality data, and updating the digital twinning model. Through the synergistic effect of digital twinning pre-screening, constrained multi-objective optimization and a model self-learning mechanism, the intelligent level, optimization efficiency and long-term adaptability of blended yarn process design are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of textile process optimization, specifically to an intelligent design system for blended yarn processes that combines digital twins and multi-objective optimization. Background Technology

[0002] Blended yarns are an important product in the textile industry. By blending fibers with different properties in a certain proportion, the advantages of each component fiber can be combined to meet the diverse needs of functional textiles. Polyester and nylon blended yarns combine the high strength and abrasion resistance of polyester with the softness and elasticity of nylon, and are widely used in sportswear, outdoor equipment and other fields. However, the traditional blended yarn process design mainly relies on experience and trial and error. The determination of process parameters requires repeated adjustments through a large number of experiments, which is not only time-consuming and labor-intensive, but also makes it difficult to ensure the stability and consistency of yarn quality. Especially when facing multiple performance requirements, how to find the best balance between yarn strength, evenness and hairiness index has become a key issue restricting the optimization of blended yarn processes.

[0003] Existing technologies use digital twins to construct virtual mapping models of physical systems, enabling the simulation of production processes in a virtual environment, prediction of product performance, and reduction of trial-and-error costs. Multi-objective optimization algorithms, such as genetic algorithms and particle swarm optimization, can find Pareto optimal solutions among multiple conflicting objectives, providing decision-makers with various trade-off options. However, existing technologies still have shortcomings. First, digital twin models are often only used for process monitoring and fault diagnosis, failing to fully utilize their pre-screening role in process optimization. Second, the multi-objective optimization process lacks consideration of equipment constraints. Third, they lack adaptive learning capabilities; when production conditions change, the predictive accuracy of digital twin models gradually decreases. Summary of the Invention

[0004] To address the technical problems mentioned in the background section, this invention proposes an intelligent design system for blended yarn processes that combines digital twins and multi-objective optimization.

[0005] Therefore, the technical solution adopted by the present invention is as follows: A smart design system for blended yarn processes that combines digital twins and multi-objective optimization. This system includes: M1: Digital twin construction module, which collects real-time process parameters and yarn quality data of the blended production line, establishes a digital twin model of the blended process based on multi-physics field coupling simulation, simulates the combination of process parameters in the virtual environment and predicts the corresponding yarn quality data, and filters out feasible parameter space that meets the quality threshold based on the predicted yarn quality data. M2: Multi-objective optimization module, which calculates the influence weight of the process parameters on yarn quality data within the feasible parameter space; establishes a multi-objective optimization function to generate a Pareto optimal solution set; and further compares the optimal solution with the current equipment state constraints to select the optimal process parameters. M3: Dynamic control module. Based on the deviation between the optimal process parameters and the real-time process parameters, it calculates the adjustment priority of the process parameters and adjusts them accordingly. At the same time, it compares the predicted yarn quality data with the collected yarn quality data. When the prediction deviation exceeds the preset threshold, it backtracks to update the simulation parameters of the digital twin model and re-selects the feasible parameter space.

[0006] Furthermore, the process parameters include the blending ratio c, the draft ratio b, and the twist coefficient a. The yarn quality data includes yarn strength (B), yarn evenness (CV), and hairiness index (H).

[0007] Furthermore, the multiphysics coupling simulation includes friction field simulation, electrostatic field simulation, and airflow field simulation. Based on the aforementioned friction field simulation, electrostatic field simulation, and airflow field simulation, the fiber motion equation is established, expressed as:

[0008] in, Indicates fiber quality; Represents the fiber position vector; Indicates frictional force; Represents electrostatic force; Indicates airflow drag force; It represents gravity.

[0009] Furthermore, the predicted yarn quality data includes predicted values ​​for yarn strength. Predicted values ​​of uniformity of the strip Predicted values ​​of feather index .

[0010] Furthermore, the influence weights are calculated using the partial derivative method, specifically through the following steps: 1) Select the center point The center point is the geometric center of the feasible parameter space, and the partial derivatives of the process parameters with respect to the yarn strength are calculated at the center point. 2) The partial derivative of the blending ratio with respect to yarn strength in the process parameters is calculated using numerical differentiation and is expressed as:

[0011] in, This represents a small disturbance in the blending ratio; simultaneously, the partial derivatives of the draft ratio and twist coefficient with respect to yarn strength are calculated. ; 3) Calculate the influence weight of process parameters on yarn strength based on the absolute value of the partial derivatives, and normalize the result so that the sum of the influence weights is 1, expressed as:

[0012]

[0013] middle, , and These represent the weights of the blending ratio, draft ratio, and twist coefficient on yarn strength, respectively; at the same time, the weights of the process parameters on yarn evenness and hairiness index are calculated.

[0014] Furthermore, the optimization objective of the multi-objective optimization function is expressed as:

[0015] in, Indicates constraints; Represents the feasible parameter space; The optimization objective of the multi-objective optimization function is solved using a non-dominated sorting genetic algorithm, specifically, 1) Generate J individuals within the feasible parameter space. The sum of all individuals constitutes the population, and each individual represents a set of process parameters. 2) For each individual, calculate the corresponding three objective function values ​​using the digital twin model and perform non-dominated ranking; 3) Using a reference point-guided selection mechanism, the association distance between an individual and a reference point is calculated, expressed as:

[0016] in, This represents the association distance between individual j and reference point k; , and Let represent the normalized objective function values ​​of individual j; Indicates the coordinates of reference point k; 4) A new generation of population is generated through selection, crossover, and mutation operations. After G generations of evolution, the Pareto optimal solution set is output. ; Equipment state constraints include the equipment's adjustability and load limitations, based on the Pareto optimal solution set. Calculated using a weighted algorithm The optimal process parameters are obtained by calculating the comprehensive quality score of each solution.

[0017] Furthermore, the indicator for adjusting the priority is expressed as follows:

[0018] in, Indicates the deviation value of process parameters; Indicators representing the priority of process parameter adjustments; The average weight of the influence of process parameters on yarn quality data is indicated; the adjustment order is determined by sorting the adjustment priority indicators according to their magnitude.

[0019] Furthermore, the prediction bias is expressed as:

[0020] in, Indicates the prediction deviation of yarn quality data; This represents yarn quality data predicted by a digital twin model based on optimal process parameters. The simulation parameters include fiber friction coefficient, fiber surface charge density, and airflow field correction coefficient. These simulation parameters are adjusted using an optimization algorithm. The objective function for parameter identification is established as follows:

[0021] Where M represents the number of data points used for identification; , and These represent the simulation parameters respectively. The model predicts the value of the m-th data point; the particle swarm optimization algorithm is used to solve for the minimum value of the objective function for parameter identification, and the updated simulation parameters are obtained. .

[0022] Compared with the prior art, the advantages of the present invention are as follows: 1. This invention uses a digital twin model to pre-simulate multiple sets of process parameters in a virtual environment and selects feasible parameter spaces that meet quality thresholds. This avoids the blind search of optimization algorithms in infeasible parameter regions and significantly improves optimization efficiency. At the same time, the high-fidelity model established based on multi-physics coupling simulation technology can accurately predict the motion trajectory and distribution state of fibers under the combined action of friction, electrostatic and airflow fields, providing a reliable basis for process optimization.

[0023] 2. When performing multi-objective optimization within the feasible parameter space, this invention calculates the influence weight of process parameters on yarn quality data, establishes a multi-objective optimization function that comprehensively considers yarn strength, evenness, and hairiness index, and compares each Pareto optimal solution with the current equipment state constraints to select the optimal process parameters that balance excellent performance and feasibility, ensuring that the optimization results can be effectively applied in actual production.

[0024] 3. This invention establishes a prediction deviation-driven model self-learning mechanism. By comparing the predicted yarn quality data with the actual detected yarn quality data in real time, when the prediction deviation exceeds a preset threshold, the simulation parameters of the digital twin model are automatically updated and the feasible parameter space is re-selected. This mechanism can continuously adapt to changes in production conditions such as equipment wear, environmental changes, and raw material differences, maintaining long-term prediction accuracy and optimization effects, and significantly improving the intelligence level of blended yarn production and product quality stability. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is a flowchart of the intelligent design process for blended yarns according to the present invention; Figure 2 This is a flowchart of the digital twin construction module of the present invention; Figure 3 This is a flowchart of the multi-objective optimization module of the present invention. Detailed Implementation

[0027] To achieve the above objectives, the present invention provides an intelligent design system for blended yarn processes that combines digital twins and multi-objective optimization. Please refer to [link to relevant documentation]. Figures 1-3 The system includes: M1: Digital twin construction module. It collects real-time process parameters and yarn quality data of the blended production line, establishes a digital twin model of the blending process based on multiphysics coupling simulation, simulates the combination of process parameters in a virtual environment and predicts the corresponding yarn quality data, and filters out feasible parameter spaces that meet the quality threshold based on the predicted yarn quality data. The digital twin building module is responsible for collecting production data in real time, establishing virtual simulation models, and providing feasible parameter ranges for subsequent optimization through a pre-screening mechanism, thereby avoiding blind optimization and improving the efficiency and accuracy of process design.

[0028] Process parameters include blending ratio, draft ratio, and twist coefficient, which directly determine the final performance of the blended yarn. The blending ratio is monitored in real time by the feeding device of the blending machine. The feeding device accurately measures the feeding ratio of polyester fiber and nylon fiber through a weighing sensor. In this embodiment, when the blending production line is set to a blending ratio of polyester to nylon of 6:4, the feeding device monitors the feeding amount of the two fibers in real time through a high-precision electronic scale. The draft ratio is obtained in real time by the encoder on the drafting device, which includes multiple sets of rollers. The draft ratio is expressed as the speed ratio between each set of rollers. By monitoring the roller speed in real time, the degree of fiber stretching can be precisely controlled, thereby affecting the yarn strength and evenness. The twist coefficient is measured jointly by the rotation speed sensor of the twisting device and the yarn delivery speed sensor. The twist coefficient reflects the degree of twisting of the yarn and is expressed as:

[0029] in, This represents the twist coefficient at time t; Indicates the twist of the yarn (twist / meter); It indicates the linear density of the yarn; proper control of the twist coefficient can balance the strength and softness of the yarn.

[0030] Yarn quality data includes yarn strength, evenness, and hairiness index, which are collected in real time by online testing equipment. Yarn strength is measured by a tensile tester, which applies tensile force to the yarn until it breaks, and records the maximum force value B at the time of breakage. Yarn evenness is monitored in real time by a yarn evenness tester. This device uses the capacitance method to measure the quality fluctuation of the yarn. Yarn evenness (coefficient of variation) is expressed as:

[0031] in, Indicates the uniformity of the strips at time t; This represents the standard deviation of the yarn quality at time t; This represents the average yarn quality at time t; the smaller the yarn evenness, the better the yarn evenness. The hairiness index is determined by a hairiness tester, which uses a photoelectric sensor to detect the number of fibers protruding from the yarn surface. The hairiness index represents the number of hairs larger than 3mm on a unit length of yarn. The lower the hairiness index, the better the surface smoothness of the yarn.

[0032] All real-time process parameters and yarn quality data are transmitted in real time via industrial Ethernet or fieldbus technology to ensure the timeliness and accuracy of data acquisition. The data acquisition frequency is set to 10 times per second to capture subtle changes in the production process.

[0033] Based on the collected real-time data, this module establishes a digital twin model of the blending process, i.e., a virtual simulation model. This model employs multiphysics coupling simulation technology, comprehensively considering the influence of inter-fiber frictional fields, electrostatic fields, and airflow fields on the blending process. Friction field simulation simulates the friction between fibers and between fibers and equipment surfaces; polyester and nylon fibers have different coefficients of friction, resulting in differentiated motion trajectories during blending, and the frictional force between fibers... Represented as:

[0034] in, Indicates the coefficient of friction between fibers; It represents the normal pressure between fibers; by simulating the distribution of frictional force between fibers under different blending ratios, the uniformity of fiber mixing can be predicted. Electrostatic field simulation considers the static charge generated in the fibers during drawing and twisting; polyester and nylon fibers have different dielectric constants, resulting in static charges of different polarities during friction, leading to attraction or repulsion between the fibers; the electrostatic potential on the fiber surface... Solving using the Poisson equation, it can be expressed as:

[0035] in, This represents the charge density on the fiber surface; It represents the vacuum dielectric constant; the distribution of the electrostatic field affects the fiber distribution state, and thus affects the yarn evenness. Airflow field simulation demonstrates the influence of airflow in the drafting zone on fiber movement. During ring spinning, a complex airflow field exists in the drafting zone; airflow velocity and direction affect fiber straightening and alignment. Solving using the Navier-Stokes equations, it can be expressed as:

[0036] in, Indicates air density; Indicates airflow pressure; This represents the aerodynamic viscosity; by solving the airflow field distribution, the airflow guidance structure of the drafting device can be optimized, thereby improving the fiber motion state.

[0037] Multiphysics coupled simulation couples the calculation results of frictional field, electrostatic field, and airflow field to comprehensively analyze the fiber's motion trajectory and distribution state under the synergistic effect of multiple fields. The core of the coupled simulation is to establish the fiber motion equation, which comprehensively considers the forces exerted on the fiber by each physical field, and is expressed as:

[0038] in, Indicates fiber quality; Represents the fiber position vector; Indicates frictional force; Represents electrostatic force; Indicates airflow drag force; This represents gravity; by solving this equation, the complete motion trajectory of the fiber during the blending process can be obtained.

[0039] Based on the established digital twin model, multiple combinations of process parameters are simulated in a virtual environment; the blending ratio is set to vary from 3:7 to 7:3 (polyester to nylon), the draw ratio to vary from 20 to 35, and the twist coefficient to vary from 300 to 450; an orthogonal experimental design method is used to generate N combinations of process parameters, where N is determined according to the number of parameter levels, and in this embodiment, N=27 is set; For each combination of process parameters, the digital twin model calculates the corresponding fiber motion trajectory and distribution state, thereby predicting yarn quality data; the predicted value of yarn strength... Calculated based on fiber packing density and interfiber bonding force, it is expressed as:

[0040] in, This represents the predicted yarn strength under the combination of the nth set of process parameters. Indicates the strong correction factor; This represents the fiber packing density under the combination of the nth set of process parameters (obtained through simulation). This indicates the strength of the first type (polyester) fiber monofilament; This indicates the strength of the monofilament of the second type (nylon) fiber; This indicates the proportion of polyester in the blend; Predicted value of strip uniformity Based on the calculation of fiber distribution uniformity, it can be expressed as:

[0041] in, This represents the predicted value of the strip uniformity under the combination of the nth set of process parameters; Indicates the stem correction factor; The variance of yarn quality under the combination of the nth set of process parameters (obtained through simulation) reflects the uniformity of fiber distribution along the yarn length. Predicted value of feather index Calculated using the number of fiber ends and fiber cohesion, it can be expressed as:

[0042] in, This represents the predicted value of the feathering index under the combination of the nth set of process parameters; Indicates the feathering correction factor; This represents the number of fiber ends under the combination of the nth set of process parameters (obtained through simulation). This represents the fiber cohesion force under the combination of the nth set of process parameters (obtained through simulation).

[0043] In this embodiment, the three-dimensional spatial distribution of fibers is calculated through multiphysics coupling simulation. For the simulation of fiber packing density, a 100×100 grid is divided on the yarn cross-section. The grid size is determined based on the yarn diameter; for example, for a 30-count yarn with a diameter of approximately 0.19 mm, the grid size is set to 2. ×2 Then, the number of fibers within each grid is counted; fibers whose center points fall within the grid are counted. The position vector of each fiber is obtained by solving the fiber motion equation, and it is determined whether the fiber is within the grid. Finally, the spatial variance of the fiber distribution is calculated, and the fiber packing density is defined as follows:

[0044] in, The standard deviation of the number of fibers within the grid; This represents the theoretical maximum standard deviation (when all fibers are clustered in one grid); The closer the value is to 1, the more uniform and dense the fiber distribution. Verification was achieved using scanning electron microscopy (SEM) images of actual yarn sections. Ten sets of yarn samples with different process parameters were prepared, cross-sections were photographed using SEM, fiber positions were extracted using image processing algorithms, the actual fiber packing density was calculated, the simulated values ​​were compared with the actual values, and the interfiber friction coefficient in the digital twin model was adjusted to ensure that the error between the two was less than 5%. For the variance of yarn quality Along the length of the yarn, a sampling section is set every 1 mm, for a total of 1000 sections (corresponding to 1 m of yarn). At each section, the number of fibers passing through and the total fiber mass are counted. The fiber mass is calculated based on the linear density and the section length. The variance of the mass of the 1000 sections is calculated, which is the variance of the yarn mass. Regarding the number of fiber ends, all fiber ends within the simulation area (1m yarn) are counted. A fiber end is defined as a fiber length shorter than the yarn length and a distance of less than 3mm from the fiber tip to the yarn surface. For each fiber, it is determined whether it is completely wrapped inside the yarn during twisting. Calculated using the fiber motion equation and cohesion force, fiber ends with insufficient cohesion force will protrude from the yarn surface, forming fuzz. The cohesion force g is calculated using the coefficient of friction between fibers and the radial pressure generated by twisting, and is expressed as:

[0045] in, This indicates the radial normal force exerted on the fiber; This indicates the length of the fiber wrapping.

[0046] Based on the predicted yarn quality data, a feasible parameter space that meets the quality thresholds is selected. The quality thresholds are set according to the product technical requirements. In this embodiment, the yarn strength threshold B is set to 15 N, the yarn evenness threshold CV is set to 12%, and the hairiness index threshold H is set to 5 hairs per 10 meters. For each combination of process parameters, it is determined whether the predicted corresponding yarn quality data meets the threshold condition. The process parameter combinations that meet the condition constitute the feasible parameter space. Through this pre-screening mechanism, the original N parameter combinations are reduced to M feasible parameter combinations. In this embodiment, after screening, M=15 feasible parameter combinations are obtained. This method avoids blindly optimizing infeasible parameters and significantly improves optimization efficiency.

[0047] M2: A multi-objective optimization module that calculates the weights of process parameters on yarn quality data within the feasible parameter space; establishes a multi-objective optimization function to generate a Pareto optimal solution set; further compares the optimal solution with the current equipment state constraints to select the optimal process parameters. This module performs refined optimization within the feasible parameter space, comprehensively considering three optimization objectives: maximizing yarn strength, optimizing yarn evenness, and minimizing hairiness index. By calculating the influence weights, the mechanism of each parameter is clarified, a multi-objective optimization function is established to form a mathematical model, a non-dominated sorting genetic algorithm is used to solve for the Pareto optimal solution set, and the feasibility is evaluated through equipment constraints. Finally, the optimal combination of process parameters that balances performance and feasibility is selected.

[0048] Within the feasible parameter space, the degree of influence of each process parameter on yarn quality data varies. By calculating the influence weight, the role of each parameter on each quality indicator is quantified. For yarn strength, the influence weights of the blending ratio c, draft ratio b, and twist coefficient a are calculated; the partial derivative method is used at the center point of the feasible parameter space. The partial derivatives of each parameter with respect to yarn strength are calculated at the point where the center point is selected as the geometric center of the feasible parameter space. The partial derivative of yarn strength with respect to the blend ratio is calculated using a numerical differentiation method. This involves making a small perturbation to the blend ratio at the center point while keeping other parameters constant, and then calculating the partial derivative using a digital twin model. and The partial derivatives are expressed as:

[0049] in, This represents a small disturbance in the blending ratio; similarly, the partial derivatives of yarn strength with respect to the draft ratio and twist coefficient are calculated using numerical differentiation methods. In this embodiment, the minute perturbation is determined based on the numerical differentiation accuracy requirements and the adjustment accuracy of the blending ratio in actual production. Specifically, with a fixed draft ratio and twist coefficient, minute perturbations of 0.01, 0.02, 0.05, and 0.1 are tested near the center of the blending ratio (5:5), and the numerical stability of the partial derivatives is calculated. When <0.01, the numerical error exceeds 5%. When the value is greater than 0.05, the linear approximation fails; considering both computational accuracy and the effectiveness of the linear approximation, the following is determined: =0.02 (i.e., a 2% change in blending ratio), applicable to any combination of blending ratios between 3:7 and 7:3.

[0050] Then, the influence weights of each parameter on yarn strength are calculated based on the absolute values ​​of the partial derivatives, and normalized so that the sum of the three is 1, expressed as:

[0051]

[0052]

[0053] in, , and These represent the weights of the influence of blending ratio, draft ratio, and twist coefficient on yarn strength, respectively. The influence weights of each parameter on yarn evenness and hairiness index were calculated using the same method. The calculation results of the influence weights provide a quantitative basis for establishing a multi-objective optimization function, clarify the mechanism of each parameter in different quality objectives, and help to understand the comprehensive impact of parameter adjustment on yarn performance.

[0054] Based on the results of the influence weight calculation, a multi-objective optimization function is established. The optimization objectives include maximizing yarn strength, minimizing yarn evenness, and minimizing the hairiness index. These three objectives are interdependent, and a Pareto optimal balance needs to be sought through a multi-objective optimization method. The multi-objective optimization problem is expressed as follows:

[0055] in, Indicates constraints; Represents the feasible parameter space; The characteristic of this multi-objective optimization problem is that there are conflicts between the objectives. For example, increasing the twist coefficient can increase yarn strength and reduce the hairiness index, but may lead to increased yarn evenness; increasing the draft ratio can improve yarn evenness, but may reduce yarn strength. Therefore, there is no single solution that simultaneously optimizes all three objectives, but rather a set of Pareto optimal solutions, each representing a different equilibrium point among the three objectives. To solve the above multi-objective optimization problem, the Non-Dominated Sorting Genetic Algorithm (NSGA-III) is adopted; this algorithm searches for Pareto optimal solutions in the feasible parameter space by simulating natural selection and genetic mechanisms. During algorithm initialization, J individuals are randomly generated within the feasible parameter space. The sum of all individuals constitutes the population, and each individual represents a set of process parameters. For each individual, three objective function values ​​are calculated using a digital twin model, followed by non-dominated sorting to divide the population into different non-dominated levels. The first non-dominated level contains individuals not dominated by any other individual, the second non-dominated level contains individuals dominated only by individuals in the first level, and so on. The condition for individual i to dominate individual j is... and and And at least one inequality sign is strictly true; Based on non-dominated ranking, a reference point-guided selection mechanism is adopted. K reference points are uniformly distributed in the target space. For each non-dominated level, the association distance between an individual and a reference point is calculated. Individuals with a large crowding distance and association with a reference point are preferentially selected. The association distance between an individual and a reference point is expressed as:

[0056] in, This represents the association distance between individual j and reference point k; , and Let represent the normalized objective function values ​​of individual j; Indicates the coordinates of reference point k. A new generation of population is generated through selection, crossover, and mutation operations. The crossover operation uses simulated binary crossover with a crossover probability set to 0.9, and the mutation operation uses polynomial mutation. After G generations of evolution, the algorithm outputs the Pareto optimal solution set. The solutions in the Pareto optimal set represent different combinations of process parameters, achieving different balances among the three optimization objectives. To select the optimal process parameters, it is necessary to compare the optimal solutions based on the current equipment state constraints. Equipment state constraints include the equipment's adjustability and load limitations; when adjusting the blending ratio in the mixing device of the blending machine, there are limitations on the adjustment range and accuracy; the roller speed ratio adjustment of the drafting device is limited by motor performance; the spindle speed adjustment of the twisting device is limited by mechanical inertia; for each solution in the Pareto optimal solution set, the deviation from the current real-time process parameters is calculated, and it is determined whether the deviation is within the equipment's adjustability range. Solutions that meet the determination conditions are considered feasible under the current equipment state, forming an feasible solution set. ; In this embodiment, the determination of the equipment's adjustability is based on the equipment technical manual and actual debugging experience. For the blending ratio, an electronic dispensing system is used, and the range of a single adjustment of the blending ratio is [range missing]. The adjustment accuracy is ±0.5%. For example, if the current blending ratio is 6:4 (polyester to nylon), the adjustable range is 55%-65% polyester. If this range is exceeded, the machine needs to be stopped to adjust the feed rollers. Adjustment time exceeding 30 minutes is not suitable for online optimization. This parameter was obtained through performance testing of three blending machines (Jingwei E33, Rieter K44, and Murata 861) and is applicable to automatic batching systems. For the roller speed ratio of the drafting device, limited by the servo motor response speed and mechanical transmission accuracy, the single adjustment range is ±10% of the current value, and the adjustment accuracy is ±0.5%. For example, if the current draft ratio is 30, the adjustable range is 27-33. This parameter was obtained through dynamic response testing of 5 drafting systems (3-roller and 4-roller configurations) under the condition that the yarn linear speed is 80-120 m / min. For the spindle speed of the twisting device, due to the limitations of the frequency converter performance and mechanical inertia, the single adjustment range is ±15% of the current speed, and the adjustment accuracy is ±1%. For ring spinning machines, the spindle speed range is usually 8000-15000 rpm. The twist coefficient is related to the spindle speed and yarn linear speed, and the single adjustment range is ±12% of the current value. This parameter was obtained by testing the acceleration and deceleration performance of ring spinning machines (models FA506 and JWF1562).

[0057] Further considering equipment load limitations, when the equipment is operating under high load, significant parameter adjustments may lead to equipment failure or product quality fluctuations; therefore, an equipment load factor is defined. The result is calculated by comprehensively monitoring motor current, vibration signals, and temperature signals, and is expressed as follows:

[0058] in, , and These represent the current motor current, vibration amplitude, and equipment temperature, respectively. , and These represent the rated current, vibration limit, and temperature limit, respectively. , and These represent the weight coefficients of each indicator. In this embodiment, they are determined using the Analytic Hierarchy Process (AHP) and expert scoring. Five equipment engineers were invited to conduct pairwise comparisons of the importance of the three indicators: current, vibration, and temperature. A judgment matrix was constructed, and the eigenvectors were calculated to obtain the weights. After a consistency check, the weights were determined. =0.5、 =0.3、 =0.2, current reflects the power load of the equipment and is the primary indicator; vibration reflects the mechanical condition and is a secondary indicator; temperature has a lag effect and is an auxiliary indicator; when When the equipment is considered to be under high load, the solution with the smaller adjustment range is preferred. The smaller the comprehensive index of the adjustment range, the less disturbance the implementation of the solution will cause to the equipment. In the feasible solution set In this process, a weighted algorithm is used to calculate the comprehensive quality score of each solution. The comprehensive quality score takes into account the achievement of the three optimization objectives and the comprehensive index of adjustment range, and the solution with the highest comprehensive quality score is selected as the optimal process parameters.

[0059] M3: Dynamic control module. Based on the deviation between the optimal process parameters and the real-time process parameters, it calculates the adjustment priority of the process parameters and adjusts them accordingly. Simultaneously, it compares the predicted yarn quality data with the collected yarn quality data. When the prediction deviation exceeds a preset threshold, it updates the simulation parameters of the digital twin model and re-selects feasible parameter spaces. This module is responsible for applying the optimal process parameters to actual production, achieving a smooth transition by calculating and adjusting priorities, and realizing model self-learning through predictive deviation monitoring to form a complete closed-loop control.

[0060] Based on the optimal process parameters output by the multi-objective optimization module and the real-time process parameters, calculate the deviation values ​​of each process parameter. , and The adjustment priority of each process parameter is determined based on a combination of influence weight and deviation value. Parameters with high influence weight and large deviation value are adjusted first to quickly improve yarn quality. The adjustment priority index is defined as follows:

[0061]

[0062]

[0063] in, , and These are priority indicators representing the blending ratio, draft ratio, and twist coefficient, respectively. , and These represent the average influence weights of the blending ratio, draft ratio, and twist coefficient on the three quality indicators, respectively. The adjustment priority is determined by ranking the priority indicators according to their magnitude, with parameters having higher priority indicators being adjusted first.

[0064] After the process parameters are adjusted, the predicted yarn quality data is compared with the actual tested yarn quality data to evaluate the prediction accuracy of the digital twin model. The prediction deviation is expressed as:

[0065]

[0066]

[0067] in, , and These represent the prediction deviations for yarn strength, yarn evenness, and hairiness index, respectively. , and These represent the yarn quality data predicted by the digital twin model based on optimal process parameters. When any prediction deviation exceeds a preset threshold, the simulation parameters of the digital twin model are considered to need updating. The simulation parameters of the digital twin model are updated retrospectively using a parameter identification method. The simulation parameters in the digital twin model include fiber friction coefficient, fiber surface charge density, and airflow field correction coefficient. The initial values ​​of these parameters are based on literature data or experimental measurements, but may change in actual production due to factors such as equipment status and environmental conditions. In this embodiment, the preset threshold for yarn strength is determined through comparative analysis; production data from 50 batches are collected, with each batch containing 100 test points; then, the deviation distribution between the initial predicted value and the measured value of the digital twin model is calculated, and the average deviation of yarn strength is statistically obtained. and standard deviation According to 3 In principle, 99.7% of the deviation should be within... Within the range, considering that a 5% strength fluctuation is acceptable in actual production, the preset threshold for yarn strength is set at 8%. When the accuracy is >8%, the model's prediction accuracy is considered to have decreased. The preset threshold for uniformity of the stripe is a relative indicator, and its test repeatability is good. Based on statistical analysis of 50 batches of data, it was set at 10%. The feather test results show significant dispersion, therefore the threshold is relatively lenient; based on statistical analysis of 50 batches of data, it was set at 15%. When any of the three deviations exceeds the threshold, a model update is immediately triggered. This OR logic ensures a timely response to any inaccuracy in any quality metric.

[0068] Adjust simulation parameters by optimizing algorithms. To minimize the deviation between the model's predicted values ​​and the actual measured values; establish the objective function for parameter identification, expressed as:

[0069] Where M represents the number of data points used for identification; , and These represent the simulation parameters respectively. The model predicts the value of the m-th data point. The particle swarm optimization algorithm is used to solve for the minimum value of the above objective function, and the updated simulation parameters are obtained. The particle swarm optimization algorithm searches for the optimal solution in the parameter space by simulating the foraging behavior of bird flocks; the number of particles is set to 30, the maximum number of iterations is 100, and the inertia weight is linearly decreased from 0.9 to 0.4.

[0070] After updating the simulation parameters, the digital twin model is rerun to simulate multiple sets of process parameters in a virtual environment and predict the corresponding yarn quality data. The feasible parameter space that meets the quality threshold is then re-screened. The updated feasible parameter space may be different from the original feasible parameter space, reflecting the re-evaluation of the feasibility of process parameters after the model accuracy is improved. The updated feasible parameter space is passed to the multi-objective optimization module, triggering a new round of optimization process. This prediction bias-driven model update mechanism enables the digital twin model to learn itself, continuously adapt to changes in production conditions, and maintain prediction accuracy and optimization effect.

[0071] This invention proposes an intelligent design system for blended yarn processes that combines digital twins and multi-objective optimization. Through digital twin pre-screening, constrained multi-objective optimization, and prediction bias-driven model self-learning, it optimizes the process design of blended yarns, solving problems such as high trial-and-error costs, low optimization efficiency, and poor adaptability in traditional process design methods. Specifically, a high-fidelity digital twin model is established through multi-physics coupled simulation, pre-screening feasible parameter spaces in a virtual environment to avoid blindly optimizing infeasible parameters. Multi-objective optimization considering equipment constraints improves the feasibility of the process plan while ensuring product quality. Prediction bias monitoring and model parameter backtracking updates enable the system to achieve self-learning capabilities, allowing it to continuously adapt to changes in production conditions.

[0072] In summary, by introducing digital twin technology and intelligent optimization algorithms, this invention not only improves the efficiency and accuracy of blended yarn process design and reduces trial-and-error costs and production adjustment time, but also enhances the system's adaptability and robustness, significantly improving the level of intelligence in blended yarn production and the stability of product quality.

[0073] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A smart design system for blended yarn processes combining digital twins and multi-objective optimization, characterized in that: The system includes: M1: Digital twin construction module, which collects real-time process parameters and yarn quality data of the blended production line, establishes a digital twin model of the blended process based on multi-physics field coupling simulation, simulates the combination of process parameters in the virtual environment and predicts the corresponding yarn quality data, and filters out feasible parameter space that meets the quality threshold based on the predicted yarn quality data. M2: Multi-objective optimization module, which calculates the influence weight of the process parameters on yarn quality data within the feasible parameter space; establishes a multi-objective optimization function to generate a Pareto optimal solution set; and further compares the optimal solution with the current equipment state constraints to select the optimal process parameters. M3: Dynamic control module. Based on the deviation between the optimal process parameters and the real-time process parameters, it calculates the adjustment priority of the process parameters and adjusts them accordingly. At the same time, it compares the predicted yarn quality data with the collected yarn quality data. When the prediction deviation exceeds a preset threshold, it updates the simulation parameters of the digital twin model and re-screens the feasible parameter space.

2. The intelligent design system for blended yarn processes combining digital twins and multi-objective optimization as described in claim 1, characterized in that, The process parameters include the blending ratio c, the draft ratio b, and the twist coefficient a. The yarn quality data includes yarn strength (B), yarn evenness (CV), and hairiness index (H).

3. The intelligent design system for blended yarn processes combining digital twins and multi-objective optimization as described in claim 2, characterized in that, The multiphysics coupling simulation includes friction field simulation, electrostatic field simulation, and airflow field simulation. Based on the aforementioned friction field simulation, electrostatic field simulation, and airflow field simulation, the fiber motion equation is established, expressed as: in, Indicates fiber quality; Represents the fiber position vector; Indicates frictional force; Represents electrostatic force; Indicates airflow drag force; It represents gravity.

4. The intelligent design system for blended yarn processes combining digital twins and multi-objective optimization as described in claim 3, characterized in that, The predicted yarn quality data includes predicted values ​​for yarn strength. Predicted values ​​of uniformity of the strip Predicted values ​​of feather index .

5. The intelligent design system for blended yarn processes combining digital twins and multi-objective optimization as described in claim 4, characterized in that, The influence weights are calculated using the partial derivative method, and the specific steps are as follows: 1) Select the center point The center point is the geometric center of the feasible parameter space, and the partial derivatives of the process parameters with respect to the yarn strength are calculated at the center point. 2) The partial derivative of the blending ratio with respect to yarn strength in the process parameters is calculated using numerical differentiation and is expressed as: in, This represents a small disturbance in the blending ratio; simultaneously, the partial derivatives of the draft ratio and twist coefficient with respect to yarn strength are calculated. ; 3) Calculate the influence weight of process parameters on yarn strength based on the absolute value of the partial derivatives, and normalize the result so that the sum of the influence weights is 1, expressed as: middle, , and These represent the weights of the blending ratio, draft ratio, and twist coefficient on yarn strength, respectively; at the same time, the weights of the process parameters on yarn evenness and hairiness index are calculated.

6. The intelligent design system for blended yarn processes combining digital twins and multi-objective optimization according to claim 5, characterized in that, The optimization objective of the multi-objective optimization function is expressed as: in, Indicates constraints; Represents the feasible parameter space; The optimization objective of the multi-objective optimization function is solved using a non-dominated sorting genetic algorithm, specifically, 1) Generate J individuals within the feasible parameter space. The sum of all individuals constitutes the population, and each individual represents a set of process parameters. 2) For each individual, calculate the corresponding three objective function values ​​using the digital twin model and perform non-dominated ranking; 3) Using a reference point-guided selection mechanism, the association distance between an individual and a reference point is calculated, expressed as: in, This represents the association distance between individual j and reference point k; , and Let represent the normalized objective function values ​​of individual j; Indicates the coordinates of reference point k; 4) A new generation of population is generated through selection, crossover, and mutation operations. After G generations of evolution, the Pareto optimal solution set is output. ; Equipment state constraints include the equipment's adjustability and load limitations, based on the Pareto optimal solution set. Calculated using a weighted algorithm The optimal process parameters are obtained by calculating the comprehensive quality score of each solution.

7. The intelligent design system for blended yarn processes combining digital twins and multi-objective optimization according to claim 6, characterized in that, The indicator for adjusting the priority is expressed as follows: in, Indicates the deviation value of process parameters; Indicators representing the priority of process parameter adjustments; The average weight of the influence of process parameters on yarn quality data is indicated; the adjustment order is determined by sorting the adjustment priority indicators according to their magnitude.

8. The intelligent design system for blended yarn processes combining digital twins and multi-objective optimization according to claim 7, characterized in that, The prediction bias is expressed as: in, Indicates the prediction deviation of yarn quality data; This represents yarn quality data predicted by a digital twin model based on optimal process parameters. The simulation parameters include fiber friction coefficient, fiber surface charge density, and airflow field correction coefficient. These simulation parameters are adjusted using an optimization algorithm. Establish the objective function for parameter identification, expressed as: Where M represents the number of data points used for identification; , and These represent the simulation parameters respectively. The model predicts the value of the m-th data point; the particle swarm optimization algorithm is used to solve for the minimum value of the objective function for parameter identification, and the updated simulation parameters are obtained. .

Images