Method for analyzing and optimizing quality fluctuation of multiple correlation parameters in spinning process

Abstract

The application discloses a method for analyzing and optimizing quality fluctuation of multiple associated parameters in a spinning process, and specifically comprises the following steps: step 1, predicting the spinning process multiple associated parameters and yarn quality; step 2, performing sensitivity analysis on the spinning process parameters by using a variance-based sensitivity analysis method, and screening out key spinning process parameters that have a greater impact on the yarn quality; and step 3, modeling the quality optimization of the multiple associated parameters in the spinning process. According to the method, the sensitivity of the spinning process parameters to the yarn quality is analyzed by determining the forward relationship between the spinning process parameters and the yarn quality, the key spinning process parameters are screened out, the quality optimization model of the spinning process parameters is established, the optimal spinning process parameter combination is obtained by solving the quality optimization model of the multiple associated parameters in the spinning process, and guidance for improving the yarn quality is provided.

Description

Technical Field

[0001] This invention belongs to the field of quality control technology in spinning processes, specifically relating to a method for analyzing and optimizing the quality fluctuations of multiple related parameters in the spinning process. Background Technology

[0002] The rapid development and application of information technologies such as industrial big data, cloud computing, and artificial intelligence have had a disruptive impact on spinning production activities. Digital intelligent monitoring systems can perceive, collect, and monitor big data information in real time, including raw cotton fiber properties, machine operating status, and yarn quality indicators during the spinning process. This data contains a wealth of information closely related to yarn quality. Therefore, leveraging spinning data to maintain quality stability and provide rapid and reasonable optimization suggestions plays a crucial role in improving yarn quality. In recent years, the sheer volume of quality-related information collected during the spinning process, coupled with the interrelationships among these process parameters, has led to inaccurate mappings between these parameters and yarn quality indicators. Traditional methods, relying solely on experience to determine spinning process parameters, have limitations; overly conservative experience can result in inappropriate parameter selection. Parameter optimization through experimental verification requires repeated experiments, resulting in significant time lags. This hinders the spinning process quality detection system from quickly responding to anomalies and taking timely corrective measures. Therefore, establishing a practical and feasible spinning process parameter optimization model is of significant value for improving yarn quality and increasing production efficiency. Domestic and foreign scholars have conducted extensive research on the optimization of spinning process parameters from different perspectives and have provided solutions: Shao Jingfeng et al. linearized the relationship between fiber property variables and yarn quality characteristic values, established a feedback mechanism between process parameters and yarn breaking strength, and constructed a process parameter optimization method based on fuzzy multi-criteria. [Shao Jingfeng, Ma Chuangtao. A Smart Control Model for Spinning Quality Based on Multi-Process Knowledge Association [J]. Control Theory and Applications, 2018, 35(06): 840-849]. Yang Jianguo et al. used genetic algorithm to optimize the weights and thresholds of gradient backpropagation neural network and established a forward model of spinning process parameters. On this basis, they used genetic algorithm to invert the process parameters of combed wool spinning production process. [Yang Jianguo, Xiong Jingwei, Xu Lan et al. Inversion Model of Combed Wool Spinning Process Parameters Using Hybrid Population Genetic Neural Network [J]. Journal of Textile Research, 2016, 37(07): 149-154]. Based on the teaching-learning-based optimization algorithm, Diyaley et al. determined the optimal parameter combination of front and rear zone variables and input variables in the ring spinning process, and produced yarn with quality indicators that met the requirements [DIYALEY S, CHAR ABORTYS. Teaching-learning-based optimization of ring and rotor spinning processes[J].Soft Computing, 2021, 25(15):10287-10307].Based on the performance characteristics of Ula grass fiber, Sun Ying et al. optimized the parameters of each pre-spinning process by pre-treating Ula grass fiber, adopting a weighing raw material mixing process, and optimizing the parameters of combing roller speed, rotor process and spinning tension in rotor spinning process [Sun Ying, Kong Lingping, Xiang Xianxin et al. Process optimization of Ula grass blended rotor yarn [J]. Cotton Textile Technology, 2018, 46(10): 44-46.].

[0003] Existing research on the optimization of spinning process parameters mainly focuses on optimizing spinning process parameters, without providing a set of optimization schemes for multiple related parameters in the spinning process. However, in the actual spinning process, the properties of raw cotton fibers and spinning process parameters jointly determine the final yarn quality. The collected data on quality-related spinning process parameters is enormous, and these spinning process parameters influence each other, which makes the optimization model for spinning process parameters complex and increases the difficulty of solving the optimization model. Summary of the Invention

[0004] The purpose of this invention is to provide a method for analyzing and optimizing the quality fluctuations of multiple related parameters in the spinning process. By determining the forward relationship between spinning process parameters and yarn quality, the sensitivity of spinning process parameters to yarn quality is analyzed, key spinning process parameters are screened, a quality optimization model for spinning process parameters is established, and the optimal combination of spinning process parameters is obtained by solving the quality optimization model of multiple related parameters in the spinning process, thus providing guidance for improving yarn quality.

[0005] The technical solution adopted in this invention is a method for analyzing and optimizing the quality fluctuations of multiple related parameters in the spinning process, specifically including the following steps:

[0006] Step 1: Predicting the relationship between multiple parameters in the spinning process and yarn quality;

[0007] Step 2: Use the variance-based sensitivity analysis method to perform sensitivity analysis on the spinning process parameters and screen out the key spinning process parameters that have a significant impact on yarn quality.

[0008] Step 3: Modeling for quality optimization of multiple related parameters in the spinning process.

[0009] The invention is further characterized in that,

[0010] Step 1 is implemented in the following steps:

[0011] Step 1.1: Given a spinning process parameter set X = [x1, x2, ... x] consisting of m-dimensional spinning process parameters. i ,…,x m A yarn quality dataset Y = [y1, y2, ..., y] composed of p-dimensional yarn quality indices. j ,…y pThe spinning process parameter X is input into the GRNN model through the input layer. In the multi-parameter quality prediction model of the spinning process, the number of neurons in the input layer is equal to the dimension m of the spinning process parameter set.

[0012] Step 1.2: The number of neurons in the pattern layer of the GRNN model is equal to the number of training samples, n. The transfer function of the pattern layer neurons is:

[0013]

[0014] In the formula, k = 1, 2, ..., n, and X is the network input variable; X k Let σ be the learning sample corresponding to the k-th neuron, and σ be the width coefficient of the Gaussian function, which is used to adjust the sensitivity of the neuron to the input layer. When σ is small and close to 0, the predicted value is closer to the training sample value, and the model has poor generalization ability. When σ is set to a large value, the predicted value is close to the mean of the test set samples.

[0015] Step 1.3: The summation layer of the GRNN model uses two types of neurons S. D With S j One method involves summing the outputs of all pattern layer neurons in the GRNN model arithmetically, resulting in its transfer function. Another approach involves a weighted summation of the outputs of all neurons in the pattern layer. The connection weight between the i-th neuron and the j-th molecular summation neuron in the pattern layer is equal to the weight of the k-th output sample y. k The j-th element in the sequence is given by its transfer function: j = 1, 2, ..., p;

[0016] Step 1.4: Obtain the predicted value f(*) of a certain yarn quality index from the output layer of the GRNN model. j ,f(*) j It is calculated using the following formula:

[0017]

[0018] Based on the above steps, the yarn quality prediction value is obtained by predicting the spinning process parameters.

[0019] Step 2 is implemented in the following steps:

[0020] Step 2.1: Perform variance decomposition, decomposing the spinning process parameters x... i The fluctuations in yarn quality index Y caused by the fluctuations are decomposed into Var(Y) as: x i The degree of fluctuation in yarn quality caused by a single change (V) i and x i With x j The coupling effect between them causes yarn quality fluctuations Vij By analogy, we can obtain the following formula:

[0021]

[0022] In the formula, x ～i It means divided by x i A vector consisting of all other spinning process parameters, E(·) denotes the expectation operator, Var(·) denotes the variance operator, and V i To measure the spinning process parameter x i The average reduction in the variance of the output yarn quality index Y when fixed across its entire distribution domain, x ～ij Indicates division by x i With x j Other spinning process parameters, V ij To measure the spinning process parameter x i With x j The average reduction in the variance of the output yarn quality index Y is calculated by fixing it across its entire distribution domain, and so on, to calculate V. 12…m ;

[0023] Step 2.2: Define the first-order effect of the sensitivity index for the spinning process as S. i , representing x i The contribution rate of fluctuation to yarn quality fluctuation, S i Represented as:

[0024]

[0025] Besides the first-order effect, there are also second-, third-, and higher-order effects. The second-order effect represents the effect on the spinning process parameter x. i With x j The contribution rate of the coupling effect between the two to yarn quality fluctuations; the third-order effect represents the spinning process parameter x. i x j x k The contribution rate of the coupling effect between them to yarn quality fluctuations, and so on, leads to:

[0026]

[0027] Therefore, the spinning process parameter x i The total effect of fluctuation on yarn quality index Y is x. i The sum of the first-order effects caused by the spinning process and all higher-order effects caused by its coupling with other spinning process parameters is represented by S. Ti The specific expression is:

[0028]

[0029] Finally, based on the reference [MARIO P. Sensitivity analysis in practice: a guide to assessing scientific models[J]. Journal of the American Statistical Association, 2006, 101(473):398-399], approximate calculations were performed. Using SOBOR sequence, H×m sample matrices A and B are generated by random sampling between the upper and lower limits of the spinning process parameters. H is selected as 200 according to the actual situation; m H×m matrices are established. make The i-th column of matrix B is equal to the i-th column of matrix B, and such that... Matrix A is equal to matrix B except for column i. Then, matrices A, B, and C are... As input, f(A) is calculated using spinning process parameters and a yarn quality prediction model. g f(B) g , Further obtain variance operators and The expectation operator is expressed as follows:

[0030]

[0031] In equation (7), f(*) g Let g represent the g-th regression value in the sample matrix, where g = 1, 2, ..., H, and obtain the total order sensitivity S of the spinning process parameters. Ti The parameters of each spinning process are sorted and their contribution to the overall sensitivity are calculated. The top k parameters with a sum of contributions of more than 80% are selected as key spinning process parameters to establish a quality optimization model for spinning process parameters, where k < m.

[0032] In step 2.2, H is selected as 200.

[0033] Step 3 is implemented in the following steps:

[0034] Step 3.1: The actual spinning process is a complex and dynamic process involving multiple factors. With the application of intelligent manufacturing monitoring systems, dynamically adjusting spinning process parameters based on the current yarn quality status using monitoring system data is the simplest and most effective means to achieve effective yarn quality control. Therefore, the mathematical model for optimizing spinning process parameters is established as follows:

[0035] Obj min|Y-Y'|

[0036]

[0037] Here, Y' represents the ideal value of the yarn quality index, minx i With maxx i These are the minimum and maximum values ​​of parameters used in the spinning process;

[0038] The spinning process involves numerous parameters, leading to computational redundancy and local optima when solving multi-parameter optimization models. The sparrow search algorithm, based on the foraging and danger-avoidance behaviors of sparrows, exhibits rich population diversity. Therefore, this invention utilizes the sparrow search algorithm to solve the multi-parameter quality optimization model for the spinning process.

[0039] Step 3.2: Solve the multi-parameter quality optimization model of the spinning process using the sparrow search algorithm, specifically as follows:

[0040] The solution process of the multi-parameter quality optimization model in the spinning process can be abstracted into an explorer-follower-watcher model, which iteratively updates the group's foraging location by superimposing a reconnaissance and early warning mechanism.

[0041] First, the key spinning process parameter constraint minx i With maxx i Randomly generate an initial population sample D of size G within the range. g1 ,x g2 ,…,x gk g = 1, 2, ..., G.

[0042] Then, spinning parameters are optimized. During the search process, explorers will prioritize obtaining food and have a larger foraging search range. Their position is updated as follows:

[0043]

[0044] here, The position of the i-th sparrow in the j-th dimension at iteration number t corresponds to the key spinning process parameter combination at time t; α is a random number between (0,1); iter max The maximum number of iterations; Q ~ N(u,σ) 2 L is a 1×d matrix with all elements equal to 1; R2∈[0,1] is the warning value; ST is the safety value (0.5,1]. When R2<ST, it means there is no danger around the population, and a global search is performed. When R2≥ST, it means a predator has been found, and all sparrows need to change their search direction. During foraging, once the discoverer finds food, the followers will immediately update their positions to grab the food. The follower's position update method is as follows:

[0045]

[0046] Here, D bThe optimal position found for the current individual sparrow, i.e., the current optimal combination of key spinning process parameters; D Worst The worst-case position is the combination of critical spinning process parameters in the global worst-case scenario; A + =A T (AA T ) -1 Elements in A are randomly assigned the value -1 or 1. This means that when i > n / 2, the i-th follower with a lower fitness value needs to adjust its hunting direction, and vice versa;

[0047] The vigilant is responsible for detecting threats. When a predator is detected, it abandons its current location and flies to a new location. The vigilant's location is updated as follows:

[0048]

[0049] Here, D Best The current global optimal position represents the current global optimal combination of key spinning process parameters; β is a random number representing the step size, β ~ N(0,1); u is a random number; f i This represents the current sparrow fitness value, which is the absolute value of the difference between the actual and ideal values ​​of the yarn quality index at the current sparrow's location; f B and f W These are the globally best and worst fitness values; For a minimal constant that makes the denominator non-zero, when f i >f B At that time, the sparrow was on the edge of the population, when f i =f B When this occurs, it indicates that the watcher has spotted a predator and needs to approach other sparrows to avoid being preyed upon.

[0050] Finally, by continuously changing the combination of spinning process parameters through the above-mentioned explorer, follower, and watchdog position update methods, and calculating the actual values ​​of yarn quality indicators, the position where the difference between the ideal and actual values ​​of yarn quality indicators is minimized is found. The key spinning process parameter combination at this position is taken as the optimal key spinning process parameter combination, providing guidance for improving yarn quality.

[0051] The beneficial effects of this invention are:

[0052] Considering the interconnected and multi-dimensional characteristics of spinning process parameters, this invention proposes a method for analyzing and optimizing the quality fluctuations of multiple interconnected parameters in the spinning process. This method combines neural networks to obtain the mathematical mapping relationship between spinning process parameters and yarn quality, establishing a predictive model between these parameters. Furthermore, it incorporates sensitivity analysis to quantify the impact of spinning process parameter fluctuations on yarn quality, identifying key spinning process parameters and simplifying the quality optimization model for multiple interconnected parameters. By establishing this model and utilizing intelligent optimization algorithms, the optimal combination of spinning process parameters is obtained. This invention provides a reference solution for quality prediction, sensitivity analysis, and quality improvement of spinning process parameters. Attached Figure Description

[0053] Figure 1 This is an overall flowchart of the method of the present invention;

[0054] Figure 2 It uses GRNN to compare the predicted and actual values ​​of yarn quality;

[0055] Figure 3 These are the results of sensitivity analysis of spinning process parameters;

[0056] Figure 4 It is the overall sensitivity contribution of the spinning process parameters;

[0057] Figure 5 It is an iterative curve for optimizing the quality of parameters in the spinning process. Detailed Implementation

[0058] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0059] This invention provides a method for analyzing and optimizing the quality fluctuations of multiple related parameters in a spinning process, comprising the following three parts: based on Figure 1 As shown, firstly, to investigate the impact of subtle fluctuations in multiple related parameters during the spinning process on yarn, a GRNN-based study on the relationship between multiple related parameters and yarn quality prediction was conducted, obtaining the mathematical mapping relationship between spinning process parameters and yarn quality. Then, a sensitivity analysis based on variance was used to perform sensitivity analysis on the spinning process parameters, identifying key spinning process parameters that significantly impact yarn quality. Finally, with the objective of minimizing the absolute value of the ideal and actual values ​​of yarn quality indicators, a quality optimization model for multiple related parameters during the spinning process was constructed using the key spinning process parameters. The optimal combination of multiple related parameters was obtained through iterative solution using the sparrow search algorithm. The specific steps are as follows:

[0060] Step 1: Predicting the relationship between multiple correlation parameters in the spinning process and yarn quality.

[0061] First, it is necessary to fully explore the mapping relationship between multiple correlation parameters in spinning and yarn quality, and to deeply analyze the impact of changes in these parameters on yarn quality indicators. GRNN, a supervised neural network, possesses strong mapping capabilities and robustness, adapting to the nonlinearity and small sample size of spinning data. It can accurately and promptly reflect the impact of subtle fluctuations in multiple correlation parameters on yarn quality indicators. Based on this, this section utilizes GRNN to establish a predictive model between multiple correlation parameters in spinning and yarn quality indicators.

[0062] GRNN essentially corrects RBF neural networks through non-iterative learning, without needing to initialize network connection weights. GRNN has a four-layer structure: input layer, pattern layer, summation layer, and output layer.

[0063] Step 1.1: Given a spinning process parameter set X = [x1, x2, ... x] consisting of m-dimensional spinning process parameters. i ,…,x m A yarn quality dataset Y = [y1, y2, ..., y] composed of p-dimensional yarn quality indices. j ,…y p The spinning process parameter X is input into the GRNN model through the input layer. In the multi-parameter quality prediction model of the spinning process, the number of neurons in the input layer is equal to the dimension m of the spinning process parameter set.

[0064] Step 1.2: The number of neurons in the pattern layer of the GRNN model is equal to the number of training samples, n. The transfer function of the pattern layer neurons is:

[0065]

[0066] In the formula, k = 1, 2, ..., n, and X is the network input variable; X k Let σ be the learning sample corresponding to the k-th neuron, and σ be the width coefficient of the Gaussian function, which is used to adjust the sensitivity of the neuron to the input layer. When σ is small and close to 0, the predicted value is closer to the training sample value, and the model has poor generalization ability. When σ is set to a large value, the predicted value is close to the mean of the test set samples.

[0067] Step 1.3: The summation layer of the GRNN model uses two types of neurons S. D With S j One method involves summing the outputs of all pattern layer neurons in the GRNN model arithmetically, resulting in its transfer function. Another approach involves a weighted summation of the outputs of all neurons in the pattern layer. The connection weight between the i-th neuron and the j-th molecular summation neuron in the pattern layer is equal to the weight of the k-th output sample y. k The j-th element in the sequence is given by its transfer function: j = 1, 2, ..., p;

[0068] Step 1.4: Obtain the predicted value f(*) of a certain yarn quality index from the output layer of the GRNN model. j ,f(*) j It is calculated using the following formula:

[0069]

[0070] Based on the above steps, the yarn quality prediction value is obtained by predicting the spinning process parameters.

[0071] Step 2: Use the variance-based sensitivity analysis method to perform sensitivity analysis on the spinning process parameters and screen out the key spinning process parameters that have a significant impact on yarn quality.

[0072] In actual spinning processes, spinning parameters exhibit multidimensional coupling and nonlinearity. Optimizing spinning process parameters that significantly impact yarn quality is an effective means of improving yarn quality. Therefore, sensitivity analysis is introduced to analyze the degree to which changes in multiple related parameters of different spinning processes affect yarn quality, and to screen out the key parameters that have a significant impact on yarn quality.

[0073] Spinning parameters not only affect yarn quality individually, but also have a certain coupling effect between different spinning process parameters, and the resulting interaction affects yarn quality. Based on this, a variance-based sensitivity analysis method is used to analyze the degree of influence of spinning process parameters and their mutual coupling on yarn quality.

[0074] Step 2.1: Perform variance decomposition, decomposing the spinning process parameters x... i The fluctuations in yarn quality index Y caused by the fluctuations are decomposed into Var(Y) as: x i The degree of fluctuation in yarn quality caused by a single change (V) i and x i With x j The coupling effect between them causes yarn quality fluctuations V ij By analogy, we can obtain the following formula:

[0075]

[0076] In the formula, x ～i It means divided by x i A vector consisting of all other spinning process parameters, E(·) denotes the expectation operator, Var(·) denotes the variance operator, and V i To measure the spinning process parameter x i The average reduction in the variance of the output yarn quality index Y when fixed across its entire distribution domain, x～ij Indicates division by x i With x j Other spinning process parameters, V ij To measure the spinning process parameter x i With x j The average reduction in the variance of the output yarn quality index Y is calculated by fixing it across its entire distribution domain, and so on, to calculate V. 12…m ;

[0077] Step 2.2: Define the first-order effect of the sensitivity index for the spinning process as S. i , representing x i The contribution rate of fluctuation to yarn quality fluctuation, S i Represented as:

[0078]

[0079] Besides the first-order effect, there are also second-, third-, and higher-order effects. The second-order effect represents the effect on the spinning process parameter x. i With x j The contribution rate of the coupling effect between the two to yarn quality fluctuations; the third-order effect represents the spinning process parameter x. i x j x k The contribution rate of the coupling effect between them to yarn quality fluctuations, and so on, leads to:

[0080]

[0081] Therefore, the spinning process parameter x i The total effect of fluctuation on yarn quality index Y is x. i The sum of the first-order effects caused by the spinning process and all higher-order effects caused by its coupling with other spinning process parameters is represented by S. Ti The specific expression is:

[0082]

[0083] Finally, based on the reference [MARIO P. Sensitivity analysis in practice: a guide to assessing scientific models[J]. Journal of the American Statistical Association, 2006, 101(473):398-399], approximate calculations were performed. Using SOBOR sequence, H×m sample matrices A and B are generated by random sampling between the upper and lower limits of the spinning process parameters. H is selected as 200 according to the actual situation; m H×m matrices are established. make The i-th column of matrix B is equal to the i-th column of matrix B, and such that... Matrix A is equal to matrix B except for column i. Then, matrices A, B, and C are... As input, f(A) is calculated using spinning process parameters and a yarn quality prediction model. g f(B) g , Further obtain variance operators and The expectation operator is expressed as follows:

[0084]

[0085] In equation (7), f(*) g Let g represent the g-th regression value in the sample matrix, where g = 1, 2, ..., H. The total order sensitivity S of the spinning process parameters can then be obtained. Ti The parameters of each spinning process are sorted and their contribution to the overall sensitivity are calculated. The top k (k < m) parameters with a sum of contributions of more than 80% are selected as key spinning process parameters to establish a quality optimization model for spinning process parameters.

[0086] Step 3: Quality optimization modeling of multiple related parameters in the spinning process;

[0087] Step 3.1: The actual spinning process is a complex and dynamic process involving multiple factors. With the application of intelligent manufacturing monitoring systems, dynamically adjusting spinning process parameters based on the current yarn quality status using monitoring system data is the simplest and most effective means to achieve effective yarn quality control. Therefore, the mathematical model for optimizing spinning process parameters is established as follows:

[0088] Obj min|Y-Y'|

[0089]

[0090] Here, Y' represents the ideal value of the yarn quality index, minx i With maxx i These represent the minimum and maximum values ​​of parameters used in the spinning process.

[0091] The spinning process involves numerous parameters, leading to computational redundancy and local optima when solving multi-parameter optimization models. The sparrow search algorithm, based on the foraging and danger-avoidance behaviors of sparrows, exhibits rich population diversity. Therefore, this invention utilizes the sparrow search algorithm to solve the multi-parameter quality optimization model for the spinning process.

[0092] Step 3.2: Solve the multi-parameter quality optimization model of the spinning process using the sparrow search algorithm, specifically as follows:

[0093] The solution process of the multi-parameter quality optimization model in the spinning process can be abstracted into an explorer-follower-watcher model, which iteratively updates the group's foraging location by superimposing a reconnaissance and early warning mechanism.

[0094] First, the key spinning process parameter constraint minx i With maxx i Randomly generate an initial population sample D of size G within the range. g1 ,x g2 ,…,x gk g = 1, 2, ..., G.

[0095] Then, spinning parameters are optimized. During the search process, explorers will prioritize obtaining food and have a larger foraging search range. Their position is updated as follows:

[0096]

[0097] here, The position of the i-th sparrow in the j-th dimension at iteration number t corresponds to the key spinning process parameter combination at time t; α is a random number between (0,1); iter max The maximum number of iterations; Q ~ N(u,σ) 2 L is a 1×d matrix with all elements equal to 1; R2∈[0,1] is the warning value; ST is the safety value (0.5,1]. When R2<ST, it means there is no danger around the population, and a global search is performed. When R2≥ST, it means a predator has been found, and all sparrows need to change their search direction. During foraging, once the discoverer finds food, the followers will immediately update their positions to grab the food. The follower's position update method is as follows:

[0098]

[0099] Here, D b The optimal position found for the current individual sparrow, i.e., the current optimal combination of key spinning process parameters; D Worst The worst-case position is the combination of critical spinning process parameters in the global worst-case scenario; A + =A T (AA T ) -1 Elements in A are randomly assigned the value -1 or 1. This means that when i > n / 2, the i-th follower with a lower fitness value needs to adjust its hunting direction, and vice versa.

[0100] The vigilant is responsible for detecting threats. When a predator is detected, it abandons its current location and flies to a new location. The vigilant's location is updated as follows:

[0101]

[0102] Here, D Best The current global optimal position represents the current global optimal combination of key spinning process parameters; β is a random number representing the step size, β ~ N(0,1); u is a random number; f i This represents the current sparrow fitness value, which is the absolute value of the difference between the actual and ideal values ​​of the yarn quality index at the current sparrow's location; f B and f W These are the globally best and worst fitness values; For a minimal constant that makes the denominator non-zero, when f i >f B At that time, the sparrow was on the edge of the population, when f i =f B When this occurs, it indicates that the watcher has spotted a predator and needs to approach other sparrows to avoid being preyed upon.

[0103] Finally, by continuously changing the combination of spinning process parameters through the above-mentioned explorer, follower, and watchdog position update methods, and calculating the actual values ​​of yarn quality indicators, the position where the difference between the ideal and actual values ​​of yarn quality indicators is minimized is found. The key spinning process parameter combination at this position is taken as the optimal key spinning process parameter combination, providing guidance for improving yarn quality.

[0104] Example 1

[0105] Application Examples of Quality Fluctuation Analysis and Optimization Methods for Multiple Correlated Parameters in the Spinning Process

[0106] The effectiveness of the multi-parameter quality optimization method for the spinning process proposed in this paper is verified using production quality data of a certain combed wool yarn. The 75 sets of production quality data of combed wool yarn given in the literature include: sliver oil content ( / %), roving twist coefficient ( / %), fiber diameter ( / µm), fiber length ( / mm), fiber diameter dispersion coefficient ( / %), fiber quality unevenness ( / %), yarn draft ratio, yarn traveler number, and yarn speed (r·min). -1 Ten spinning process parameters, including yarn quality index, are used, with the yarn CV value being the yarn quality index.

[0107] 1) Screening of key parameters in the spinning process

[0108] Based on the GRNN yarn quality prediction model, 1-70 sets of samples were used as the model construction sample set, and 71-75 sets of samples were used as the model accuracy test sample set. To obtain an accurate mapping relationship between spinning process parameters and yarn quality, multiple experiments were conducted. When the smoothing coefficient σ equals 0.7, the average absolute error of the prediction model is 0.2437. Comparison of GRNN predicted values ​​with actual values ​​is shown below. Figure 2As shown, the GRNN-based yarn quality prediction model has good prediction accuracy, and the obtained prediction model can well reflect the mapping relationship between spinning process parameters and yarn quality.

[0109] The first step in the variance-based sensitivity analysis of spinning process parameters is to determine the range of values ​​for the spinning process parameters. The maximum and minimum values ​​of the spinning process parameters in the dataset are used as the range of values, as shown in Table 1. Ten 200×10 sample matrices, A and B, are generated using SOBO sequence random sampling. Based on the calculation of the first-order sensitivity and total-order sensitivity of the spinning process parameters, as follows: Figure 3 As shown, the contribution of the total order sensitivity of the spinning process parameters is as follows: Figure 4 As shown.

[0110] Table 1 Decision variables for spinning parameter optimization

[0111]

[0112] Depend on Figure 3 It can be seen that the ranking of the first-order sensitivity of each spinning process parameter is inconsistent with the analysis results of the overall sensitivity. For example, the absolute value of the first-order sensitivity of the roving twist coefficient (x2) is lower than that of the dispersion coefficients of fiber length (x5) and fiber diameter (x6), indicating that the individual effect of x2 on yarn quality is lower than that of x5 and x6. However, the overall effect of x2 is greater than that of x5 and x6, indicating that when considering the coupling relationship between x2 and other parameters, the influence of x2 on the CV value of the yarn quality index is greater than that of x5 and x6. Figure 4 The spinning process parameters are as follows: yarn traveler number x9, roving twist coefficient x2, sliver oil content x1, yarn draft ratio x8, sliver moisture regain x3, and yarn speed x. 10 The sum of the sensitivity contributions of the seven process parameters, including fiber diameter x4, is greater than 80%, indicating that these seven process parameters have a significant impact on the CV value of wool yarn. In summary, the parameters to be retained are: yarn traveler number x9, roving twist coefficient x2, sliver oil content x1, yarn draft ratio x8, sliver moisture regain x3, and yarn speed x4. 10 The spinning process parameters were optimized by using fiber diameter x4.

[0113] 2) Quality optimization of multiple related parameters in the spinning process

[0114] Based on the sensitivity analysis obtained in step 1), the spinning process parameters are: yarn traveler number x9, roving twist coefficient x2, sliver oil content x1, yarn draft ratio x8, sliver moisture regain x3, and yarn speed x. 10 Using fiber diameter x4 as the optimization variable for the spinning process parameters, and taking the median yarn CV value (%) of 19 in the dataset as the ideal value, the following optimization model for the spinning process parameters is established:

[0115]

[0116] Next, the sparrow search algorithm is used to solve the multi-parameter quality optimization model of the spinning process. The number of sparrows is set to 200, the number of explorers to 140, the number of followers to 40, the number of watchdogs to 20, the safety threshold to 0.6, the number of iterations to 100, and σ... 2 With values ​​equal to 0.3 and gam equal to 10, the quality optimization of spinning process parameters is obtained in the iterative curve as shown below. Figure 5 As shown, the optimization of spinning process parameters based on SSA converges at iteration number 11, where the optimal value is 0.0003. The optimal spinning process parameter combination is [32.159 5.2 14.8 0.8 24.540 8588.677 21.708]. Therefore, the key parameters of the spinning process can be adjusted according to the optimal spinning process parameter combination to improve yarn quality.

[0117] Example 2

[0118] A method for analyzing and optimizing quality fluctuations of multiple correlated parameters in the spinning process, specifically including the following steps:

[0119] Step 1: Predicting the relationship between multiple parameters in the spinning process and yarn quality;

[0120] Step 2: Use the variance-based sensitivity analysis method to perform sensitivity analysis on the spinning process parameters and screen out the key spinning process parameters that have a significant impact on yarn quality.

[0121] Step 3: Modeling for quality optimization of multiple related parameters in the spinning process.

[0122] Example 3

[0123] A method for analyzing and optimizing quality fluctuations of multiple correlated parameters in the spinning process, specifically including the following steps:

[0124] Step 1: Predicting the relationship between multiple parameters in the spinning process and yarn quality;

[0125] Step 1 is implemented in the following steps:

[0126] Step 1.1: Given a spinning process parameter set X = [x1, x2, ... x] consisting of m-dimensional spinning process parameters. i ,…,x m A yarn quality dataset Y = [y1, y2, ..., y] composed of p-dimensional yarn quality indices. j ,…y p The spinning process parameter X is input into the GRNN model through the input layer. In the multi-parameter quality prediction model of the spinning process, the number of neurons in the input layer is equal to the dimension m of the spinning process parameter set.

[0127] Step 1.2: The number of neurons in the pattern layer of the GRNN model is equal to the number of training samples, n. The transfer function of the pattern layer neurons is:

[0128]

[0129] In the formula, k = 1, 2, ..., n, and X is the network input variable; X k Let σ be the learning sample corresponding to the k-th neuron, and σ be the width coefficient of the Gaussian function, which is used to adjust the sensitivity of the neuron to the input layer. When σ is small and close to 0, the predicted value is closer to the training sample value, and the model has poor generalization ability. When σ is set to a large value, the predicted value is close to the mean of the test set samples.

[0130] Step 1.3: The summation layer of the GRNN model uses two types of neurons S. D With S j One method involves summing the outputs of all pattern layer neurons in the GRNN model arithmetically, resulting in its transfer function. Another approach involves a weighted summation of the outputs of all neurons in the pattern layer. The connection weight between the i-th neuron and the j-th molecular summation neuron in the pattern layer is equal to the weight of the k-th output sample y. k The j-th element in the sequence is given by its transfer function: j = 1, 2, ..., p;

[0131] Step 1.4: Obtain the predicted value f(*) of a certain yarn quality index from the output layer of the GRNN model. j ,f(*) j It is calculated using the following formula:

[0132]

[0133] Based on the above steps, the yarn quality prediction value is obtained by predicting the spinning process parameters.

[0134] Step 2: Use the variance-based sensitivity analysis method to perform sensitivity analysis on the spinning process parameters and screen out the key spinning process parameters that have a significant impact on yarn quality.

[0135] Step 3: Modeling for quality optimization of multiple related parameters in the spinning process.

Claims

1. A method for analyzing and optimizing the quality fluctuations of multiple correlated parameters in the spinning process, characterized in that, Specifically, the following steps are included: Step 1: Predicting the relationship between multiple parameters in the spinning process and yarn quality; Step 2: Use the variance-based sensitivity analysis method to perform sensitivity analysis on the spinning process parameters and screen out the key spinning process parameters that have a significant impact on yarn quality. Step 3: Quality optimization modeling of multiple related parameters in the spinning process; Step 3 is implemented in the following steps: Step 3.1: Establish a mathematical model for optimizing the spinning process parameters as follows: （8） here, This represents the ideal value for yarn quality indicators. and These are the minimum and maximum values ​​of parameters used in the spinning process; Step 3.2: Solve the multi-parameter quality optimization model of the spinning process using the sparrow search algorithm, specifically as follows: First, constraints on key spinning process parameters and Randomly generate capacity within the range G Initial population sample ; Then, spinning parameters are optimized. During the search process, explorers will prioritize obtaining food and have a larger foraging search range. Their position is updated as follows: （9） here, For the number of iterations The first time i The sparrow in the first j The location of the dimension's region, corresponding to Key spinning process parameter combinations at specific moments; for Random numbers between; This represents the maximum number of iterations. ; For elements all equal to 1 matrix; The warning value; for The safe value; when When this indicates that there is no danger around the population, a global search is performed. When a predator is detected, all sparrows need to change their foraging direction. During the foraging process, once the discoverer finds food, the followers will immediately update their positions to steal it. The follower's position update method is as follows: （10） here, The optimal position found for the current individual sparrow, i.e. the current optimal combination of key spinning process parameters; This represents the worst-case position globally, i.e., the worst-case combination of critical spinning process parameters globally. , The elements in the array are randomly assigned the values ​​-1 or 1. Meaning, when At that time, the first with a lower fitness value One follower needs to adjust its hunting direction, and vice versa; The vigilant is responsible for detecting threats. When a predator is detected, it abandons its current location and flies to a new location. The vigilant's location is updated as follows: （11） here, This represents the current global optimal position, i.e., the current global optimal combination of key spinning process parameters; The step size is a random number. ; It is a random number; This represents the current sparrow fitness value, which is the absolute value of the difference between the actual value and the ideal value of the yarn quality index at the current sparrow's position. and These are the globally best and worst fitness values; For a minimal constant that makes the denominator non-zero, when At that time, the sparrow was on the edge of the population. When this occurs, it indicates that the watcher has spotted a predator and needs to approach other sparrows to avoid being preyed upon. Finally, by continuously changing the combination of spinning process parameters through the above-mentioned explorer, follower, and watcher position update methods, and calculating the actual value of the yarn quality index, the position with the smallest difference between the ideal value and the actual value of the yarn quality index is found, and the key spinning process parameter combination at this position is taken as the optimal key spinning process parameter combination.

2. The method for analyzing and optimizing the quality fluctuation of multiple related parameters in the spinning process according to claim 1, characterized in that, Step 1 is implemented in the following steps: Step 1.1, given by Spinning process parameter set composed of spinning process parameters , Yarn quality dataset composed of yarn quality indices The spinning process parameters are input through the GRNN model. In the GRNN model, the number of neurons in the input layer is equal to the dimension of the spinning process parameter set in the multi-parameter quality prediction model of the spinning process. ; Step 1.2: The number of neurons in the pattern layer of the GRNN model is equal to the number of training samples. The transfer function of the pattern layer neurons is: （1） In the formula, , Input variables for the network; For the first k The learning samples corresponding to each neuron is the width coefficient of the Gaussian function; Step 1.3: The summation layer of the GRNN model uses two types of neurons. and One method involves summing the outputs of all pattern layer neurons in the GRNN model arithmetically, resulting in its transfer function. Another type involves a weighted summation of the outputs of all neurons in the pattern layer, where the first neuron in the pattern layer... The first neuron and the second The connection weights between the summing neurons of molecules are the first... k Output samples y k The first in j There are 4 elements, therefore its transfer function is: , ; Step 1.4: Obtain the predicted value of a certain yarn quality index from the output layer of the GRNN model. , It is calculated using the following formula: （2） Based on the above steps, the yarn quality prediction value is obtained by predicting the spinning process parameters.

3. The method for analyzing and optimizing the quality fluctuation of multiple related parameters in the spinning process according to claim 1, characterized in that, Step 2 is implemented in the following steps: Step 2.1: Perform variance decomposition to decompose the spinning process parameters. Fluctuations affect yarn quality indicators fluctuations Decomposed into: The degree of fluctuation in yarn quality caused by a single change as well as and The coupling effect between them causes yarn quality fluctuations. And so on, we get the following formula: （3） In the formula, , , Indicates by (excluding) A vector composed of other spinning process parameters besides... Represents the expectation operator. Represents the variance operator. To measure the parameters of the spinning process Output yarn quality index when fixed across its entire distribution range The average decrease in variance Indicates except and Other spinning process parameters, To measure the parameters of the spinning process and Output yarn quality index when fixed across its entire distribution range The average decrease in variance is calculated in this manner. , For the parameters of the spinning process; Step 2.2: Define the first-order effect of the sensitivity index for the spinning process as... ,express The contribution rate of fluctuation to yarn quality fluctuation. Represented as: （4） In addition to first-order effects, there are second-, third-, and higher-order effects. Second-order effects represent parameters of the spinning process. and The contribution rate of the coupling effect between the parameters to yarn quality fluctuations; the third-order effect represents the parameters of the spinning process. , , The contribution rate of the coupling effect between them to yarn quality fluctuations, and so on, leads to: （5） Therefore, spinning process parameters Fluctuations on yarn quality indicators The total effect is The sum of the first-order effects caused by the spinning process and all higher-order effects caused by its coupling with other spinning process parameters is expressed as: The specific expression is: （6） Finally, calculate The upper and lower limits of the spinning process parameters are generated by random sampling using a sobol sequence. Sample matrix and ;Establish indivual matrix ,make matrix Column equals of Column, and make Matrix division Columns and Matrices If they are equal, then the matrix will be equal. , , As input, the yarn quality prediction model is calculated using spinning process parameters. , , Further obtained variance operators and The expectation operator is expressed as follows: （7） In equation (7), Represents the first in the sample matrix One regression value, The overall sensitivity of the spinning process parameters was obtained. The parameters of each spinning process are ranked and their contribution to the overall sensitivity is calculated. The parameters with a sum of contributions of 80% or more are selected. A quality optimization model for spinning process parameters is established using these parameters as key spinning process parameters. .

4. The method for analyzing and optimizing the quality fluctuation of multiple related parameters in the spinning process according to claim 3, characterized in that, In step 2.2, Select 200.

Images