A method for predicting high-strength yarn quality

Abstract

The present invention relates to the field of yarn quality prediction, and in particular to a high-strength yarn quality prediction method, comprising: collecting images of silk threads when they are twisted, obtaining contours of each silk thread, connecting two points on the contours of the silk thread to form a reference line; taking the ratio of the product of the lateral deviation value and the longitudinal deviation value to the pre-acquired quality value as a first evaluation value; constructing a segmentation line on the image, the segmentation line intersecting with the reference line to form a plurality of reference lines; obtaining a second evaluation value based on the difference between the length of the reference line in the real-time image and the length of the reference line in the historical moment image, and obtaining a third evaluation value based on the ratio of the length of the reference line in the real-time image; inputting the first evaluation value, the second evaluation value, the third evaluation value, and the real-time collected drawing speed and silk thread tension value into a trained time series network model to predict the yarn quality. The present invention predicts the quality of the yarn through comprehensive analysis of multiple factors, thereby improving the accuracy of the yarn quality prediction.

Description

Technical Field

[0001] The invention relates to the field of yarn quality prediction, and in particular to a high-strength yarn quality prediction method. Background Art

[0002] High-strength yarn is usually used to make high-quality textiles, such as high-end clothing, car seats, aerospace materials, etc. High-strength yarn is composed of multiple silk threads. During the production process, it is necessary to use corresponding equipment to twist multiple silk threads into one strand. The quality of the silk thread twisting has a great influence on the quality of the yarn, so it is necessary to test the silk thread twisting process.

[0003] The patent application document with publication number CN115035024A discloses a yarn twist quality assessment method based on image processing. The method collects a yarn curve image in a yarn traction zone, determines the type of yarn traction abnormality, and obtains a traction abnormality evaluation index of the yarn curve in combination with the distance variance; obtains a twisted yarn image based on key points in the yarn twist zone, and obtains an overall tightness evaluation index of the twisted yarn from the grayscale value of the pixel points on the center line area in the twisted yarn image; obtains a yarn twist quality assessment index by combining the traction abnormality evaluation index and the overall tightness evaluation index, and adjusts the cradle pressure of the spinning machine in real time according to the yarn twist quality assessment index.

[0004] During the yarn twisting process, the quality of the yarn is affected by many factors, such as the tension of the silk thread, the moving speed of the silk thread, and the influence of wind in the external environment. The slight wind force causes the silk thread to stretch and deform and vibrate, thereby affecting the quality of the yarn. The above scheme only involves the evaluation of the abnormal traction of the yarn, without considering the influence of other factors during the yarn twisting, and uses a single factor to evaluate the yarn quality, resulting in low accuracy of the evaluation result. Summary of the invention

[0005] In order to solve the problem that the evaluation of yarn quality using a single factor leads to low accuracy of the evaluation result, the present invention provides a high-strength yarn quality prediction method.

[0006] The present invention provides a high-strength yarn quality prediction method, comprising:

[0007] Collecting images of silk threads when they are twisted, obtaining contour lines of silk threads in each image, and connecting two points on the contour lines of the silk threads to form a reference line, wherein the images include real-time images and images at historical moments;

[0008] Using the reference line and the contour line to obtain a transverse deviation value and a longitudinal deviation value of the thread, and taking the ratio of the product of the transverse deviation value and the longitudinal deviation value to the pre-obtained quality value as a first evaluation value;

[0009] A segmentation line is constructed on the image, and the segmentation line intersects with the reference line to form a plurality of reference lines;

[0010] A second evaluation value is obtained based on a difference between the lengths of the reference line in the real-time image and the reference line in the historical moment image, and a third evaluation value is obtained based on a ratio of the lengths of any two reference lines in the real-time image;

[0011] The first evaluation value, the second evaluation value, the third evaluation value, and the real-time collected drawing speed and wire tension value are input into the trained time series network model to predict the yarn quality;

[0012] Among them, the maximum value of the vertical distance between the point on the contour line and the baseline is taken as the lateral deviation value, and the point corresponding to the maximum distance value is recorded as the deviation point; the distance between the projection point of the deviation point on the baseline and the pre-acquired baseline point is taken as the longitudinal deviation value.

[0013] In the process of obtaining the first evaluation value, the present invention comprehensively considers the deviation of the silk thread contour caused by the influence of external force on the silk thread during twisting, and at the same time predicts the yarn quality by collecting the drawing speed and silk thread tension value during twisting and obtaining the first evaluation value, the second evaluation value and the third evaluation value. The yarn quality is predicted by multiple factors, thereby improving the accuracy of yarn quality prediction.

[0014] Preferably, the method of collecting images of the silk threads when they are twisted, obtaining the contours of each silk thread, and connecting two points on the contours of the silk thread to form a reference line is:

[0015] Convert the image to grayscale and use median filtering to remove noise;

[0016] Use Otsu thresholding to segment the grayscale image and obtain the outline of the silk thread;

[0017] The two points on the contour line that are the farthest apart are used as the starting point and the ending point of the silk line, where the ending point is located below the starting point;

[0018] Connect the starting point and the ending point to get the baseline of the thread.

[0019] The points on the contour line are used to construct a baseline to facilitate the calculation of the second evaluation value and the third evaluation value.

[0020] Preferably, the second evaluation value is calculated as follows:

[0021] Setting at least one segmentation line in the horizontal direction on the grayscale image;

[0022] The dividing line intersects with the plurality of silk threads so that the dividing line forms a plurality of first reference lines;

[0023] The first reference line in the real-time collected image corresponds to the first reference line in the historical moment image one by one, and the lengths of the two corresponding first reference lines are subtracted to obtain a plurality of difference values;

[0024] The absolute values ​​of the plurality of differences are summed up as the second evaluation value.

[0025] By comparing the real-time collected image with the image at the historical moment to obtain the second evaluation value, we can better understand the degree of deviation of the silk thread profile relative to the image at the historical moment, which is convenient for predicting the quality of the yarn during twisting.

[0026] Preferably, the third evaluation value is calculated as follows:

[0027] The calculation method of the third evaluation value is:

[0028] Two adjacent silk threads are recorded as a set of reference lines;

[0029] Two dividing lines are set in the horizontal direction on the grayscale image so that the dividing lines have intersections with each group of silk threads;

[0030] Connecting two intersection points along the direction of the dividing line or the direction of the silk thread to form a plurality of second reference lines;

[0031] Calculate the ratio of the lengths of any two second reference lines to obtain multiple ratios, calculate the difference between the multiple ratios in pairs to obtain multiple differences, record the sum of the absolute values ​​of the multiple differences as an intermediate value, and a group of silk lines corresponds to one intermediate value;

[0032] The sum of the multiple intermediate values ​​corresponding to the multiple groups of threads is recorded as the third evaluation value.

[0033] By analyzing the length of the second reference line to obtain the third evaluation value, the real-time change degree of the silk thread profile can be obtained, which is convenient for predicting the quality of the yarn during twisting.

[0034] Preferably, the method for obtaining the quality value is to perform a tension test on the wire and record the tension value corresponding to the breakage of the wire as the quality value.

[0035] By collecting the ultimate tensile force value that the silk thread can withstand, it is convenient to compare it with the tensile force value when the silk thread is twisted, thus providing a basis for predicting the quality of the yarn.

[0036] Preferably, the training method of the temporal network model is:

[0037] Obtaining a first evaluation value, a second evaluation value, a third evaluation value, a wire drawing speed, and a wire drawing force value at the same time, and setting a label for the data at the same time, wherein the label is a quality grade;

[0038] Input the acquired data and corresponding labels into the temporal network model for training;

[0039] Calculate the cross entropy loss function between the output yarn quality level and the actual yarn quality level at the next moment;

[0040] Use gradient descent to adjust model parameters to minimize the error between the predicted value and the true value;

[0041] Iteratively adjust the parameters of the time series network model until the cross entropy loss function is less than the set value or reaches the preset number of iterations to obtain a trained time series network model.

[0042] Preferably, the reference point is the midpoint of the reference line.

[0043] Preferably, the temporal network model is a long short-term memory network model.

[0044] The present invention has the following beneficial effects:

[0045] The yarn quality is predicted using a temporal network model. During the prediction process, the tension of the silk thread during the yarn twisting process, the silk thread running speed, the degree of deviation of the silk thread profile relative to the historical moment, and the real-time change degree of the silk thread profile are comprehensively considered. The quality of the yarn is predicted through a comprehensive analysis of multiple factors, which improves the accuracy of the yarn quality prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] By reading the detailed description below with reference to the accompanying drawings, the above and other purposes, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-restrictive manner, and the same or corresponding reference numerals represent the same or corresponding parts.

[0047] Figure 1 It is a method flow chart of steps S1 to S6 of a high-strength yarn quality prediction method according to an embodiment of the present invention.

[0048] Figure 2 It is a schematic diagram of yarn twisting in a high-strength yarn quality prediction method in an embodiment of the present invention.

[0049] Figure 3 It is a schematic diagram of a first state where a silk thread and a dividing line intersect in a high-strength yarn quality prediction method according to an embodiment of the present invention.

[0050] Figure 4 It is a schematic diagram of the second state of the intersection of the silk thread and the dividing line in a high-strength yarn quality prediction method according to an embodiment of the present invention.

[0051] Reference numerals: 10, silk thread; 11, yarn; 12, dividing line. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0053] It should be understood that when the terms "first", "second", etc. are used in the claims, descriptions, and drawings of the present invention, they are only used to distinguish different objects, rather than to describe a specific order. The terms "include" and "comprise" used in the description and claims of the present invention indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their collections.

[0054] The embodiment of the present invention discloses a method for predicting the quality of high-strength yarns. The specific scenario targeted by the present invention is as follows: Figure 2 As shown, during the processing of the yarn 11, a plurality of silk threads 10 are twisted together to form the yarn 11, and the quality of the yarn 11 is predicted.

[0055] The embodiment of the present invention discloses a high-strength yarn quality prediction method, referring to Figure 1 , including steps S1 to S6, which are specifically as follows:

[0056] S1. Collect images of silk threads when they are twisted, obtain contour lines of each silk thread, and connect two points on the contour lines of the silk threads to form a reference line. The images include real-time images and historical moment images.

[0057] A black background board is placed on the background area of ​​the silk thread during the twisting process to highlight the silk thread part, making it easy to extract and calculate the silk thread part on the image. Images are collected at equal intervals, and the collected images are the process of converging multiple silk threads to form a polyester yarn.

[0058] The image is converted into a grayscale image and the noise is removed by using a median filter. The region of interest is set so that the silk thread bends and falls inside the region of interest. The image of the region of interest is segmented using the Otsu threshold to obtain the silk thread contour line. The two points on the contour line that are farthest apart are used as the starting point and the ending point of the silk thread, where the ending point is located below the starting point. A straight line is used to connect the starting point and the ending point to obtain the baseline of the silk thread.

[0059] S2. Obtain a transverse deviation value and a longitudinal deviation value of the thread using the reference line and the contour line, and use the ratio of the product of the transverse deviation value and the longitudinal deviation value to the pre-obtained quality value as a first evaluation value.

[0060] Calculate the vertical distance between the point on the contour line and the reference line, record the maximum distance value as the lateral deviation value, and record the point corresponding to the maximum distance value as the deviation point; select a point on the reference line as the reference point, and the distance between the projection point of the deviation point on the reference line and the reference point is the longitudinal deviation value. The first evaluation value is the ratio of the product of the lateral deviation value and the longitudinal deviation value to the pre-obtained quality value. In this embodiment, the reference point is the midpoint of the reference line. In other embodiments, the position of the reference point can also be set based on experience.

[0061] During the yarn twisting process, the silk thread will fluctuate due to external force. The larger the deviation value, the greater the fluctuation degree of the silk thread and the greater the deformation of the silk thread, which further indicates that the operation stability of the production equipment is poor, and ultimately leads to poor yarn production quality. Conversely, it means that the operation stability of the production equipment is good and the yarn production quality is good.

[0062] S21. The method for obtaining the quality value is: extracting a silk thread from a reel, performing a tensile test on the extracted silk thread using a tensile machine, and recording the tensile force value corresponding to the breakage of the silk thread as the quality value. The greater the quality value of the silk thread, the greater the tensile force that the silk thread can withstand, and the better the quality of the yarn after twisting. On the contrary, it means that the tensile force that the silk thread can withstand is smaller, and the quality of the yarn after twisting is poor.

[0063] By taking the ratio of the product of the lateral deviation value and the longitudinal deviation value to the pre-obtained quality value as the first evaluation value and comprehensively considering multiple aspects of information such as the lateral deviation, longitudinal deviation and quality value, it is easy to obtain a more comprehensive and objective evaluation result, avoiding the one-sidedness and subjectivity brought about by a single indicator evaluation, and improving the accuracy and reliability of the evaluation.

[0064] S3. construct a segmentation line on the image, and the segmentation line intersects with the silk thread to form a plurality of reference lines.

[0065] Reference Figure 3 and Figure 4 , a segmentation line 12 is constructed on the image, and the segmentation line 12 intersects with the silk thread to form a plurality of reference lines.

[0066] S4. Obtain a second evaluation value based on a difference between the lengths of the reference line in the real-time image and the lengths of the reference line in the historical moment image.

[0067] During the twisting process of the same batch of silk threads, images of the silk threads being twisted are collected at equal intervals. The images collected at the current moment are real-time images, and the images collected before the current moment are historical moment images.

[0068] The calculation method of the second evaluation value is: at least one dividing line is set in the horizontal direction on the grayscale image, and the dividing line intersects with multiple silk threads so that the dividing line forms multiple sub-segments, and the sub-segments are recorded as first reference lines; multiple first reference lines are formed in sequence along the direction of the dividing line; the first reference line in the real-time image is matched one by one with the first reference line in the historical moment image, and the lengths of the two corresponding first reference lines are subtracted to obtain multiple difference values; the absolute values ​​of the multiple difference values ​​are summed as the second evaluation value.

[0069] For example, Figure 3 As shown, for the sake of convenience, it is assumed that the number of silk threads is 3, the dividing line and the silk threads in the real-time acquired image intersect to form a first reference line ab and a first reference line bc, the length of the first reference line ab in the real-time image is 6, the length of the first reference line bc is 7, the length of the first reference line ab in the historical moment image is 5, the length of the first reference line bc is 4, then the second evaluation value is |6-5|+|7-4|=4.

[0070] S5. Obtain a third evaluation value based on a ratio of lengths of any two reference lines in the real-time image.

[0071] The calculation method of the third evaluation value is as follows: two adjacent baselines are recorded as a group of baselines, and there are multiple groups of baselines in the grayscale image; two dividing lines are set in the horizontal direction on the grayscale image, so that the dividing lines have intersections with each group of baselines; two intersections are connected along the dividing line direction or the silk thread direction to form multiple second reference lines; the ratio of the lengths of any two second reference lines is calculated to obtain multiple ratios, and the multiple ratios are subtracted from each other to obtain multiple differences, and the sum of the absolute values ​​of the multiple differences is recorded as an intermediate value, and one group of baselines corresponds to one intermediate value; the sum of the multiple intermediate values ​​corresponding to the multiple groups of baselines is recorded as the third evaluation value.

[0072] For example, Figure 4 As shown, in the real-time collected image, two segmentation lines intersect with a group of silk lines, and their intersection points are A, B, C, D, and E. Connect two of the intersection points to form a line segment. , , , , , ,So:

[0073] By line segment For example, we get multiple ratios: , , , , , then the absolute value of the difference between these ratios is: , , , …, ;

[0074] Similarly, we can get the line segment , , , , Take the absolute value of the difference of the ratios corresponding to the example;

[0075] The absolute values ​​of these differences are further added to obtain the median value of a group of silk threads.

[0076] S6. Input the first evaluation value, the second evaluation value, the third evaluation value, and the real-time collected drawing speed and wire tension value into the trained time series network model to predict the yarn quality.

[0077] The training method of the time series network model is as follows: obtaining the first evaluation value, the second evaluation value, the third evaluation value, the drawing speed and the wire tension value at the same time, setting a label for the data at the same time, the label is the quality grade, in this embodiment, the quality grade is good or bad, in other embodiments, the quality grade can also be excellent, medium or poor, and the quality grade is set according to the actual situation; inputting the obtained data and the corresponding label into the time series network model for training; calculating the cross entropy loss function between the output yarn quality grade and the actual wire quality grade at the next moment; adjusting the model parameters by the gradient descent method to minimize the error between the predicted value and the true value; iteratively adjusting the parameters of the time series network model until the cross entropy loss function is less than the set value or reaches the preset number of iterations, and obtaining the trained time series network model. The time series network model is a long short-term memory network model.

[0078] The first evaluation value, the second evaluation value, the third evaluation value, and the real-time collected drawing speed and wire tension values ​​are input into the trained time series network model. The time series network model outputs the yarn quality grade at the next moment to predict the yarn quality. If the predicted yarn quality at the next moment is good, the yarn production continues; otherwise, the production is stopped and the equipment is inspected.

[0079] The scheme of the present invention comprehensively considers the silk thread tension value, the drawing speed and the deviation degree of the silk thread in the twisting process when predicting the quality of the yarn, and compares the real-time image of the silk thread with the image at a historical moment. It refers to multiple factors in multiple dimensions at the same time, and comprehensively predicts the quality of the yarn, thereby improving the accuracy of predicting the quality of the yarn.

[0080] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

[0081] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A high-strength yarn quality prediction method, characterized in that: include: Collecting images of silk threads when they are twisted, obtaining contour lines of silk threads in each image, and connecting two points on the contour lines of the silk threads to form a reference line, wherein the images include real-time images and images at historical moments; The lateral deviation value and the longitudinal deviation value of the silk thread are obtained by using the reference line and the contour line, and the ratio of the product of the lateral deviation value and the longitudinal deviation value to the pre-obtained quality value is used as the first evaluation value; the quality value is obtained by performing a tensile test on the silk thread, and recording the tensile value corresponding to the breakage of the silk thread as the quality value; A segmentation line is constructed on the image, and the segmentation line intersects with the silk thread to form a plurality of reference lines; A second evaluation value is obtained based on a difference between the lengths of the reference line in the real-time image and the reference line in the historical moment image, and a third evaluation value is obtained based on a ratio of the lengths of any two reference lines in the real-time image; The first evaluation value, the second evaluation value, the third evaluation value, and the real-time collected drawing speed and wire tension value are input into the trained time series network model to predict the yarn quality; The maximum vertical distance between the point on the contour line and the reference line is taken as the lateral deviation value, and the point corresponding to the maximum distance value is recorded as the deviation point; The distance between the projection point of the deviation point on the reference line and the pre-acquired reference point is taken as the longitudinal deviation value.

2. A high-strength yarn quality prediction method according to claim 1, characterized in that: The method of collecting images of the silk threads when they are twisted, obtaining the contours of each silk thread, and connecting two points on the contours of the silk thread to form a reference line is as follows: Convert the image to grayscale and use median filtering to remove noise; Use Otsu thresholding to segment the grayscale image and obtain the outline of the silk thread; The two points on the contour line that are the farthest apart are used as the starting point and the ending point of the silk line, where the ending point is located below the starting point; Connect the starting point and the ending point to get the baseline of the thread.

3. A high-strength yarn quality prediction method according to claim 2, characterized in that: The calculation method of the second evaluation value is: Setting at least one segmentation line in the horizontal direction on the grayscale image; The dividing line intersects with the plurality of silk threads so that the dividing line forms a plurality of first reference lines; The first reference line in the real-time collected image corresponds to the first reference line in the historical moment image one by one, and the lengths of the two corresponding first reference lines are subtracted to obtain a plurality of difference values; The absolute values ​​of the plurality of differences are summed up as the second evaluation value.

4. A high-strength yarn quality prediction method according to claim 2, characterized in that: The calculation method of the third evaluation value is: Two adjacent silk threads are recorded as a group of silk threads; Two dividing lines are set in the horizontal direction on the grayscale image so that the dividing lines have intersections with each group of silk threads; Connecting two intersection points along the direction of the dividing line or the direction of the silk thread to form a plurality of second reference lines; Calculate the ratio of the lengths of any two second reference lines to obtain multiple ratios, calculate the difference between the multiple ratios in pairs to obtain multiple differences, record the sum of the absolute values ​​of the multiple differences as an intermediate value, and a group of silk lines corresponds to one intermediate value; The sum of the multiple intermediate values ​​corresponding to the multiple groups of threads is recorded as the third evaluation value.

5. A high-strength yarn quality prediction method according to claim 1, characterized in that: The training method of the temporal network model is: Obtaining a first evaluation value, a second evaluation value, a third evaluation value, a wire drawing speed, and a wire drawing force value at the same time, and setting a label for the data at the same time, wherein the label is a quality grade; Input the acquired data and corresponding labels into the temporal network model for training; Calculate the cross entropy loss function between the output yarn quality level and the actual yarn quality level at the next moment; Use gradient descent method to adjust model parameters to minimize the error between predicted value and true value; Iteratively adjust the parameters of the time series network model until the cross entropy loss function is less than the set value or reaches the preset number of iterations to obtain a trained time series network model.

6. A high-strength yarn quality prediction method according to claim 1, characterized in that: The reference point is the midpoint of the reference line.

7. A high-strength yarn quality prediction method according to claim 1, characterized in that: The temporal network model is a long short-term memory network model.

Images