Method for regulating and controlling fiber net structure

Abstract

The invention relates to a method for regulating and controlling a fiber net structure, and belongs to the technical field of fiber preparation. The regulation and control method of the fiber net-shaped structure comprises the following steps: S1, carrying out image acquisition on fibers obtained by flash-jet spinning, converting the obtained image into a binary image, and calculating the ratio of the number of white particles to the total area of the image to obtain the grid density; s2, performing logistic regression model training by taking the grid density obtained in the step S1 and the flash-jet spinning parameters as training samples; and S3, predicting the grid density of the nanofiber by adopting the model trained in the step S2. According to the method, the mathematical relationship between the flash-jet spinning parameters and the fiber grid density is established by utilizing a mathematical tool, the grid density of the flash-jet spinning fibers can be predicted according to the trained model, then the adjustment of the fiber net structure is realized, a mathematical basis can be provided for the barrier property of the fibers, and a characteristic material can be conveniently produced.

Description

Technical Field

[0001] The invention belongs to the technical field of fiber preparation, and particularly relates to a method for regulating a fiber network structure. Background Art

[0002] High-pressure flash spinning is a key technology for producing high-performance fibers and their associated nonwoven materials. It is a type of solution spinning based on the principle of high-speed phase separation of a polymer solution. The fiber-forming polymer forms a spinning fluid within a high-temperature, high-pressure reaction vessel and is released under atmospheric pressure. The spinning fluid is ejected from the spinneret at high speed under high pressure and is stretched at high speed. Due to the sudden drop in pressure, the spinning fluid undergoes high-speed phase separation, and the low-boiling-point solvent and supercritical gas rapidly expand, solidifying to form continuous nanofibers with a cross-linked network structure. Currently, after fiber formation, mesh density is a key criterion for evaluating fiber barrier properties. The mesh density of a mesh-structured material generally refers to the number of skeletons or structural units within a specific volume. Mesh density affects the material's mechanical properties, such as strength, toughness, and elastic modulus. Generally, higher mesh density increases the material's strength and stiffness, but can also affect its plasticity.

[0003] Since the fiber formation process in high-pressure flash spinning technology is affected by pressure, temperature and spinning solution concentration, it is difficult to control the grid density of the high-pressure flash-sprayed mesh nanofiber bundles during spraying, which is not conducive to the production of nanofibers with specific properties.

[0004] Therefore, establishing a mathematical relationship between reaction temperature, pressure, spinning solution concentration and high-pressure flash-sprayed fiber mesh density is beneficial to regulating the barrier properties of the produced fibers. Summary of the Invention

[0005] The purpose of the present invention is to overcome the problems existing in the above-mentioned prior art and provide a method for regulating fiber network structure.

[0006] The present invention is achieved through the following technical solutions:

[0007] The present invention provides a method for regulating a fiber network structure, comprising the following steps:

[0008] S1. Capturing images of the flash-spinning fibers, converting the images into binary images, and calculating the ratio of the number of white particles to the total image area to obtain the grid density.

[0009] S2. The grid density and flash spinning parameters obtained in step S1 are used as training samples for logistic regression model training;

[0010] S3. Use the model trained in step S2 to predict the mesh density of nanofibers.

[0011] The present invention provides a mathematical relationship between the mesh density of high-pressure flash-jet fibers and flash-jet spinning parameters. The method first converts a fiber image into a binary image, i.e., a black-and-white image. After conversion, the fiber portion appears white against a black background, and the image is composed of black and white particles. The mesh density is calculated by analyzing the ratio of the number of white particles to the total area of ​​the image. A logistic regression model is then used to establish the relationship between the flash-jet spinning parameters and the mesh density, thereby enabling the prediction and regulation of the fiber mesh density. This also provides a mathematical basis for regulating the barrier properties of fibers during production.

[0012] Preferably, in step S1, Image J processing is used to obtain the number of white particles.

[0013] Specifically, the process of converting the fiber image into a binary image using ImageJ and calculating the grid density based on the ratio of the number of white particles to the total area of ​​the image can be divided into the following steps:

[0014] Step 1: Start ImageJ and open the fiber image;

[0015] Step 2: If the original image is in color, convert it to a grayscale image first;

[0016] Step 3: Adjust the threshold for binarization

[0017] In the Threshold Settings window, adjust the Min and Max sliders to ensure that the fiber part appears white and the background part appears black, and the white fibers are clearly separated from the black background. Apply;

[0018] Optionally, proceed to step 4: Image Cleanup

[0019] If there are noise or irregular areas in the binarized image, you can use the hole filling function to fill small holes in the fibers to ensure that the fibers are partially connected; you can also adjust the fiber width or remove impurities;

[0020] Step 5: Calculate the total area of ​​the image and the number of white particles

[0021] Measure total area: get the total area of ​​the image through analysis and calculation;

[0022] Measure the number of white particles: Select an appropriate particle size range to filter noise (for example, set the size range based on the size of the fiber), display the particle outline, and thus obtain the number of white particles;

[0023] Step 6: Calculate the mesh density

[0024] The grid density is obtained by the ratio of the number of white particles to the total area of ​​the image.

[0025] The formula is as follows:

[0026]

[0027] The binarized image can separate the background from the fiber foreground (white particles), simplifying subsequent analysis. Through software analysis, independent particles can be effectively identified from the image and the number of white particles can be obtained. The grid density is obtained by calculating the ratio of the number of white particles to the total area of ​​the image.

[0028] Preferably, step S2 includes data preprocessing, model building, model training and model evaluation in sequence.

[0029] Preferably, the data preprocessing includes standardizing or normalizing the flash spinning parameters.

[0030] Standardizing or normalizing the flash spinning parameters (input variables) can improve the performance of the model.

[0031] Preferably, the model establishment includes: selecting a multi-classification logistic regression model, taking grid density as the dependent variable and flash spinning parameters as the independent variables.

[0032] Preferably, in step S2, the flash spinning parameters include temperature, pressure and spinning solution concentration.

[0033] The present invention found that temperature, pressure and spinning solution concentration among the flash spinning parameters are the key factors affecting the fiber grid density. By establishing a relationship model between grid density and temperature, pressure and spinning solution concentration, the prediction effect of the model is improved.

[0034] Preferably, the spinning solution comprises a polymer, a solvent and a gas medium.

[0035] Preferably, the polymer includes at least one of polyethylene, polypropylene, polyester, polyvinylidene fluoride, cellulose acetate, and polyphenylene sulfide.

[0036] Preferably, the gas medium includes one of nitrogen, carbon dioxide, helium and argon.

[0037] Preferably, the solvent includes at least one of 1,2-dichloroethane, dichloromethane, n-pentane, and cyclohexane.

[0038] Preferably, the temperature is 120°C-220°C, the pressure is 5MPa-20MPa, and the spinning solution concentration is 1.2g / cm 3 -1.8g / cm 3 .

[0039] Preferably, the model parameters of the logistic regression model are estimated by minimizing a loss function (such as cross entropy loss).

[0040] The cross entropy loss function is:

[0041]

[0042] Among them, y i ∈{0,1}: the true label of the ii-th sample, p i : The probability that the model predicts a high density (1) grid density.

[0043] The model formula for logistic regression (for each category) is usually:

[0044]

[0045] Where (y) is the target category, (X) is the input feature (temperature, pressure, spinning solution concentration), (β k ) is the regression coefficient for each category, and (K) is the total number of categories.

[0046] Preferably, after the logistic regression model is trained, the trained model is evaluated using a test sample to obtain a trained model; the evaluation indicators include at least one of confusion matrix, precision, recall rate and F1 score.

[0047] Preferably, in the trained model, the regression coefficient of temperature is 0.05; and / or the regression coefficient of pressure is 3.4615; and / or the regression coefficient of spinning solution concentration is 0.6923.

[0048] In a second aspect, the present invention provides a spinning fiber material, including a fiber structure prepared by the method for controlling the fiber network structure.

[0049] The present invention has the following beneficial effects: the present invention uses mathematical tools to establish a mathematical relationship between flash jet spinning parameters and fiber mesh density, can predict the mesh density of flash jet spun fibers based on a trained model, and then adjust the fiber mesh structure, can provide a mathematical basis for the barrier properties of the fibers, and facilitate the production of special materials. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 Schematic diagram of a binary image of a fiber obtained by flash spinning according to an embodiment of the present invention. DETAILED DESCRIPTION

[0051] To better illustrate the purpose, technical solutions and advantages of the present invention, the present invention will be further described below in conjunction with specific embodiments. Those skilled in the art should understand that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0052] Unless otherwise specified, the experimental methods used in the examples are conventional methods; the materials, reagents, etc. used are all available from commercial sources unless otherwise specified.

[0053] Example 1

[0054] An embodiment of the present invention provides a method for regulating a fiber network structure, comprising the following steps:

[0055] Polyethylene, dichloromethane and carbon dioxide are prepared to a density of 1.3-1.6 g / cm 3 The spinning solution is prepared by using a flash spray reactor as an apparatus, and spinning is carried out at a temperature of 120-220°C and a pressure of 5-20 MPa. The spinning solution is ejected from the apparatus, and after the fiber is formed, it is placed under a microscope for observation and photography. The image of the fiber obtained by flash spray spinning is collected, and the obtained image is imported into Image J to be converted into a binary image. After Image J processing, the ratio of the number of white particles to the total area of ​​the image is calculated to obtain the grid density; the schematic diagram of the binary image is shown as follows Figure 1 As shown; the steps for image processing using ImageJ are as follows:

[0056] Step 1: Start ImageJ and open the fiber image;

[0057] Step 2: Convert the image to grayscale image;

[0058] Step 3: Adjust the threshold for binarization

[0059] In the threshold setting window, adjust the Min and Max sliders to ensure that the fiber part appears white and the background part appears black. You can adjust it in real time by observing the image until the white fibers are clearly separated from the black background. After the adjustment is completed, apply the binarization.

[0060] Step 4: Perform image cleanup (optional)

[0061] If there are noise or irregular areas in the binarized image, you can use the hole filling function to fill the small holes in the fiber to ensure that the fiber is partially connected. You can also use the image processing function to adjust the fiber width or remove impurities.

[0062] Step 5: Calculate the total area of ​​the image and the number of white particles

[0063] Measure total area: In the measurement result window, you can see the total area of ​​the image;

[0064] Measure the number of white particles: Select Analyze Particles in the Analyze menu to open the settings window. In the window, select an appropriate particle size range to filter noise (for example, set the Size range based on the size of the fiber). Select Display Results and check Show Outlines or Show Mask to display the particle outlines on the image. After clicking OK, the Results window will display the number of each particle and other statistics.

[0065] The number of white particles can be obtained through the Count value in the Results window.

[0066] Step 6: Calculate the mesh density

[0067] The grid density is obtained by the ratio of the number of white particles to the total area of ​​the image.

[0068] The formula is as follows:

[0069]

[0070] The obtained grid density and the corresponding flash spinning parameters temperature, pressure, and spinning solution concentration are then processed by Python and used as training samples for logistic regression model training to establish the relationship between temperature, pressure, spinning solution concentration and grid density. The specific steps are as follows:

[0071] Data preprocessing: A sample size of at least 10 times the number of features is required, and the input variables (temperature, pressure, spinning solution concentration) in the training samples are standardized or normalized to improve the performance of the model;

[0072] Model establishment: A multinomial logistic regression model was selected, with grid density as the dependent variable (target variable) and temperature, pressure, and spinning solution concentration as the independent variables (input variables). A random forest regression model was used, revealing that the relationship was saturated at a temperature of 220°C and a pressure of 8 MPa, so a multinomial logistic regression model was used.

[0073] Training model and model evaluation: The model formula for logistic regression (for each category) is usually:

[0074]

[0075] Where (y) is the target category, (X) is the input feature (temperature, pressure, spinning solution concentration), (β k ) is the regression coefficient for each category, (K) is the total number of categories;

[0076] Use training set and test set division: 80% of the samples are used as training samples for training, and 20% of the samples are used as test samples for testing. Perform cross-validation and divide the data set into K subsets (such as 5 or 10). Use K-1 subsets each time to train the model, and use the remaining 1 subset as the validation set. Repeat this cycle K times and finally take the average of the K results to evaluate the model to better assess the generalization ability of the model.

[0077] Use the test sample to evaluate the accuracy and predictive ability of the model, and calculate the mean square error to evaluate the classification effect;

[0078] Mean Squared Error (MSE)

[0079]

[0080] Where n is the number of samples, y i is the true value of the i-th sample, is the predicted or estimated value of the i-th sample.

[0081] Comparing MSE with the variance of the data, for example, the data below shows a mean square error of 1.71 and a variance of 12.51. This is a significant difference, so we can conclude that the model is accurate.

[0082] After Python processing, the regression coefficients are as follows: temperature: 0.05, pressure: 3.4615, spinning solution concentration: 0.6923;

[0083] Prediction: Use the trained model to predict new spinning parameters and output the predicted results of mesh density.

[0084] The predicted mesh density of the obtained model and the actual mesh density are shown in Table 1.

[0085] Table 1 Actual grid density and predicted grid density

[0086]

[0087]

[0088] As can be seen from Table 1, the error of the flash-spun fibers predicted by the method of the present invention is small, and the mesh density of the flash-spun fibers can be predicted by the flash-spinning parameters, which facilitates understanding of the fiber properties.

[0089] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the scope of protection of the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the essence and scope of the technical solutions of the present invention.

Claims

1. A method for regulating a fiber network structure, characterized in that: The following steps are involved: S1. Capturing images of the flash-spinning fibers, converting the images into binary images, and calculating the ratio of the number of white particles to the total image area to obtain the grid density. S2. The grid density and flash spinning parameters obtained in step S1 are used as training samples for logistic regression model training; S3. Use the model trained in step S2 to predict the mesh density of nanofibers.

2. The method for controlling a fiber network structure according to claim 1, characterized in that: In the step S1, Image J is used to process and obtain the number of white particles.

3. The method for controlling the fiber network structure according to claim 1, characterized in that: The step S2 includes data preprocessing, model building, model training and model evaluation in sequence.

4. The method for controlling the fiber network structure according to claim 3, characterized in that: The data preprocessing includes: standardizing or normalizing the flash spinning parameters; and / or, the model establishment includes: selecting a multi-classification logistic regression model with grid density as the dependent variable and the flash spinning parameters as the independent variable.

5. The method for controlling a fiber network structure according to claim 1, wherein: In step S2, the flash spinning parameters include temperature, pressure and spinning solution concentration.

6. The method for controlling the fiber network structure according to claim 5, characterized in that: The temperature is 120°C-220°C; and / or the pressure is 5MPa-20MPa; and / or the spinning solution concentration is 1.2g / cm 3 -1.8g / cm 3 .

7. The method for controlling the fiber network structure according to claim 5, characterized in that: The spinning solution includes a polymer, a solvent and a gas medium.

8. The method for controlling the fiber network structure according to claim 7, characterized in that: The polymer includes at least one of polyethylene, polypropylene, polyester, polyvinylidene fluoride, cellulose acetate, and polyphenylene sulfide; and / or the gas medium includes one of nitrogen, carbon dioxide, helium, and argon; and / or the solvent includes at least one of 1,2-dichloroethane, dichloromethane, n-pentane, and cyclohexane.

9. The method for controlling a fiber network structure according to claim 1, characterized in that: In step S2, the model parameters of the logistic regression model are estimated by minimizing the loss function.

10. A spinning fiber material, characterized in that: The invention comprises a fiber structure prepared by the method for controlling the fiber network structure according to any one of claims 1 to 9.

Images