Production line automatic control system based on knitted fabric processing

Abstract

The invention relates to the technical field of automatic control, and discloses a production line automatic control system based on knitted fabric processing. Comprising an observation point planning module used for planning observation points distributed at intervals, an equipment data acquisition module used for acquiring comprehensive equipment data, a first model prediction module used for predicting the equipment state of the next observation point, a quality data acquisition module used for acquiring comprehensive quality data, a second model prediction module used for predicting the equipment state of the next observation point, and a third model prediction module used for predicting the equipment state of the next observation point. The quality state prediction module is used for predicting the quality state of a next observation point, and the production processing control module is used for controlling automatic processing of a knitted fabric production line; according to the method, the processing state of the knitted fabric production line at the future moment can be predicted in advance from the two aspects of equipment and the knitted fabric, the limitation problem existing in single-dimension and single-model prediction is avoided, and the accuracy of predicting the processing state of the knitted fabric production line at the future moment is improved; and the automatic control effect of the knitted fabric production line is finally realized.

Description

Technical Field

[0001] The present invention relates to the field of automatic control technology, and more specifically, to an automatic control system for a production line based on knitted fabric processing. Background Art

[0002] The knitted fabric production line is a production line used to realize knitted fabric spinning, weaving, dyeing, finishing, post-processing and other related processing operations. It contains a large number of specific equipment with different functional types. In order to ensure that the knitted fabric production line can maintain a good processing state and improve the processing quality of knitted fabrics, it is necessary to accurately evaluate the actual processing state of the knitted fabric production line in different time periods and perform automated control operations based on the evaluation results. The patent application with reference publication number CN119002439A discloses a control system for an automated production line, including a pre-production evaluation module for scanning various types of parts to be processed in any batch, and determining the corresponding initial production scheduling target according to the scanning results, a production operation module, which includes various production line nodes, and any production line node is used to process each part to be processed in the current batch according to the initial arrangement number and output the target parts that have been processed, a capacity monitoring module, which is used to monitor the processing transition at the production line node with a preset monitoring cycle, and a production scheduling optimization module, which is used to update the arrangement order of parts at any production line node according to the update strategy and the similarity of the furnace temperature of adjacent target parts; When predicting the processing status at different times, the existing production line automation control system collects the real-time operation data of the equipment and combines the machine learning model to predict the processing status at different times to achieve the automatic control effect of the production line. However, this method has the following shortcomings: the collection method of equipment operation data will result in a single data dimension provided to the production line for comprehensive processing status prediction, which makes the single-dimensional data have limitations in prediction. At the same time, the prediction method using a single model cannot effectively predict the actual processing status of the two different aspects of the equipment and knitted fabrics in the production line, which leads to low prediction accuracy of the final production line processing status.

[0003] In view of this, the present invention proposes an automatic control system for a production line based on knitted fabric processing to solve the above problems. Summary of the invention

[0004] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solution: an automatic control system for a production line based on knitted fabric processing, applied to a control server, comprising: The observation point planning module is used to collect the extreme value of the fluctuation duration and the fluctuation value of the knitted fabric production line after the data collection point, calculate the interval duration, and plan the observation points of the interval distribution; The equipment data acquisition module is used to divide the knitted fabric production line into a feeding section, a knitting section and a winding section. Based on the full coverage principle, the comprehensive equipment data of the knitted fabric production line at the observation point is collected. The comprehensive equipment data includes the feeding speed value, the vibration drop value, the regional tension and the winding circle value; The comprehensive coverage criteria are: the source sections of the comprehensive equipment data include the feeding section, knitting section and winding section; A first model prediction module is used to predict the equipment state of the knitted fabric production line at the next observation point by using a pre-trained first prediction model; The quality data acquisition module is used to perform visual image analysis on knitted fabrics and collect comprehensive quality data of the knitted fabric production line at the observation point. The comprehensive quality data includes unit density value, local flatness and boundary width value; A second model prediction module is used to predict the quality status of the knitted fabric production line at the next observation point by using a pre-trained second prediction model; The production and processing control module is used to combine and analyze the equipment status and quality status of the same observation point, determine the production and processing status of the knitted fabric production line at the future observation point, and control the automated processing of the knitted fabric production line.

[0005] Furthermore, the collection process of the extreme value of the fluctuation duration and the fluctuation value is as follows: The period between the data collection point and the current time is recorded as the observation period, and the A data sets recorded in the observation period are queried through the database, and the sample data in the A data sets are arranged in sequence according to the order of recording; The changed sample data is recorded as valid data, and the recording times of two adjacent valid data in the A data sets are queried one by one through the timestamp, and the duration between the recording times of the two adjacent valid data is recorded as the change duration; After eliminating the maximum value of the change duration, the maximum value of the remaining change duration is recorded as the extreme value of the fluctuation duration; The digital features of two adjacent valid data are extracted respectively, the difference between the two digital features is taken and the absolute value is obtained to obtain the change amplitude, and the number of change amplitudes greater than the calibrated change value is counted to obtain the fluctuation value.

[0006] Furthermore, the marking process of the observation points is as follows: The standard duration of the knitted fabric production line is pre-set, and the interval duration is calculated based on the standard duration, the extreme value of the fluctuation duration and the fluctuation value; The calculation formula for the interval duration is: ; In the formula, is the interval duration, is the standard duration, is the extreme value of the fluctuation duration, is the fluctuation value; Take an interval as the planning interval, take the data collection point as the first observation point, and plan B non-adjacent observation points on the timeline.

[0007] Furthermore, the collection process of the vibration drop value is as follows: All the moments within B observation points are marked one by one to obtain C observation moments, and the vibration value of the knitting machine head at the C observation moments is monitored in real time by a vibration sensor installed on the knitting machine head. After removing the maximum and minimum values ​​of the vibration value, the remaining C-2 vibration values ​​are recorded as available vibration values; After subtracting the maximum available vibration value from the minimum available vibration value, B vibration drop values ​​are obtained.

[0008] Furthermore, the collection process of the winding circle value is as follows: A camera installed on the winding machine frame is used to shoot a bird's-eye view video of the winding shaft in real time, and an image of the bird's-eye view video at the first observation time of B observation points is intercepted to obtain an initial image; The axis center and axis edge of the reel are identified in the initial image by computer vision technology, a side point is randomly marked on the side edge, the axis center and the side point are connected by a straight line to obtain the center-edge line, and the initial position of the center-edge line in the initial image is identified; Identify the real-time position of the center edge line in the overhead video. When the real-time position of the center edge line coincides with the initial position once, increase the number of rotations of the reel by 1. From the first observation moment of B observation points to the last observation moment, the number of rotations of the reel in the overhead video is counted to obtain B first-circle values; Intercept the image of the last observation time of the overhead video at B observation points to obtain the termination image, and mark the real-time position of the center edge of the termination image; Along the rotation direction of the reel, measure the angle from the real-time position of the center edge of the end image to the initial position, record it as the rotation angle, and divide the rotation angle by the angle of one rotation to obtain B second circle values; After sequentially adding the B first-circle values ​​to the B second-circle values, B winding-circle values ​​are obtained.

[0009] Furthermore, the device status includes a normal status and an abnormal status; The prediction process of normal state and abnormal state is as follows: When the output of the first prediction model is 0, the predicted equipment state of the next observation point of the knitted fabric production line is normal; When the output of the first prediction model is 1, the predicted equipment state of the next observation point of the knitted fabric production line is an abnormal state.

[0010] Furthermore, the collection process of unit density value is as follows: The observation time at the middle position of the B observation points is recorded as the target time, a bird's-eye view image of the knitted fabric at the target time is captured by a camera, the bird's-eye view image is converted into a grayscale image, all the pixels in the B grayscale images are marked one by one, and the pixel values ​​of all the pixels are identified; Mark the four corner points of the grayscale image, connect the four corner points diagonally to form two diagonal lines, measure the lengths of the two diagonal lines respectively, and record one seventh of the minimum length value as the target length; Take the intersection of the two diagonals as the center of the circle and the target length as the radius to draw a central circle. Draw an outer circle with the same area as the central circle in a ring structure outside the central circle, and then sum up the central circle and the outer circle to obtain pixel circle; The pixels whose pixel values ​​are lower than the calibration pixel threshold are recorded as target pixels, and the pixels are counted one by one. The number of target pixels within the pixel circle is obtained target value; Will After dividing the target value by the total number of pixels in the pixel circle, we get The density of the sub- The density of each sub-unit is accumulated and averaged to obtain B unit density values.

[0011] Furthermore, the local flatness collection process is as follows: At the target moment of B observation points, the side view image of the knitted fabric is captured by a camera, and the location of the knitted fabric is identified by computer vision technology to obtain the knitted area; After drawing lines at the positions of the upper and lower edges of the knitted area in the horizontal direction, draw parallel upper and lower boundary lines; Find out the design thickness of the knitted fabric, keep the lower boundary line still, and adjust the distance from the upper boundary line to the lower boundary line in the vertical direction until the distance from the upper boundary line to the lower boundary line is consistent with the design thickness, and then stop adjusting; The number of pixels above the upper boundary line and the number of pixels between the upper boundary line and the lower boundary line are counted respectively to obtain a convexity value and a flatness value. After dividing the convexity value and the flatness value, B local flatness values ​​are obtained.

[0012] Further, the quality state includes a high quality state and a low quality state; The prediction process of high-quality state and low-quality state is as follows: When the output of the second prediction model is 2, the quality state of the next observation point of the knitted fabric production line is predicted to be a high quality state; When the output of the second prediction model is 3, the predicted quality state of the next observation point of the knitted fabric production line is a low quality state.

[0013] Furthermore, the production and processing state includes a shutdown state and a maintenance state, and the process of determining the shutdown state and the maintenance state is as follows: Combine and analyze the predicted equipment status and quality status of the same observation point in the future; When the equipment status and quality status of the same observation point in the future are respectively normal and high-quality, the production and processing status is determined to be the maintenance status; When the equipment status and quality status of the same observation point in the future are not abnormal and low-quality respectively, the production and processing status is determined as the shutdown status; The control process of the knitted fabric production line is as follows: When the production and processing state of the knitted fabric production line at the future Eth observation point is the maintenance state, the control terminal controls the knitted fabric production line to continue processing the knitted fabric at the Eth observation point; When the production and processing state of the knitted fabric production line at the Eth observation point in the future is a shutdown state, the control terminal controls the knitted fabric production line to stop processing the knitted fabric at the E-1th observation point.

[0014] The technical effects and advantages of the automatic control system of a production line based on knitted fabric processing of the present invention are as follows: The present invention can provide a time limit for the collection of good and bad evaluation data of the actual processing status of the knitted fabric production line by setting observation points distributed at intervals, ensure that the relevant data collected subsequently can comprehensively represent the different time points of the knitted fabric production line, and ensure that the collected data are independent. At the same time, by separately collecting the equipment data and the knitted fabric data in the knitted fabric production line, the processing status of the knitted fabric production line can be represented from two aspects of equipment and knitted fabric, and combined with two independently running prediction models, the actual status of the knitted fabric production line equipment and knitted fabric at future observation points can be predicted in advance, and the comprehensive processing status of the knitted fabric production line can be comprehensively and accurately predicted through the actual status of the equipment and knitted fabric at future observation points, thereby achieving the effect of early prediction of the processing status of the knitted fabric production line at a future moment, and further, before the knitted fabric production line is about to have abnormal failure phenomena, effective solution guidance can be provided in advance and accurately for the problems existing in the knitted fabric production line, and finally the automatic control effect of the knitted fabric production line can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1A schematic diagram of the architecture of an automated control system for a production line based on knitted fabric processing provided in the first embodiment of the present invention; Figure 2 A schematic flow chart of a production line automation control method based on knitted fabric processing provided in the second embodiment of the present invention. DETAILED DESCRIPTION

[0016] 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 only 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 ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0017] Example 1: Please refer to Figure 1 As shown, the production line automation control system based on knitted fabric processing described in this embodiment is applied to a control server, including: The observation point planning module determines the data collection points of the knitted fabric production line, collects the interval parameters of the knitted fabric production line, calculates the interval duration, and plans non-adjacent observation points on the timeline; Knitted fabric production line refers to a combination of equipment that can realize a series of complete operations of knitted fabric processing, such as knitting raw material feeding, conveying, knitting, dyeing and winding, so as to complete the whole set of production and processing flow of knitted fabric; The data collection point is the starting time for collecting various operating parameters in the knitted fabric production line, which can provide a starting limit on the timeline for subsequent diversified data collection to ensure that the various operating parameters of the knitted fabric production line can remain relatively stable after the data collection point; The operating parameters include but are not limited to the feed motor speed, preheater temperature, winding motor speed, fuel heating temperature, etc., which can represent the basic parameters of each knitting production and processing equipment involved in the knitted fabric production line.

[0018] In summary, in order to ensure that the operating parameters of the knitted fabric production line can remain relatively stable after the data collection point, it is necessary to compare the operating parameters with the pre-set standard values. The standard values ​​refer to the rated sizes of the pre-set equipment parameters. Therefore, when determining the data collection point, the control terminal is used to query the standard time when each operating parameter is equal to the standard value, and the last standard time is recorded as the data collection point in chronological order.

[0019] Interval parameters refer to the parameters that can affect the time length of the change of the diversified data on the knitted fabric production line to be sufficient to be measured, so as to provide accurate data support for the calculation of the interval time, so that the interval time can be used as the transition interval standard for the subsequent collection of various data on the knitted fabric production line; The interval parameters include the extreme value of fluctuation duration and the fluctuation amount; the extreme value of fluctuation duration and the fluctuation amount are used to respectively measure the maximum value of the duration of fluctuation of the parameters in the knitted fabric production line and the number of fluctuation parameters; The collection process of the extreme value of fluctuation duration and fluctuation amount is as follows: The period between the data collection point and the current time is recorded as the observation period, and the A data sets recorded in the observation period are queried through the database, and the sample data in the A data sets are arranged in sequence according to the order of the records; a data set refers to a collection of a large number of data at different times, so that different types of data can be comprehensively recorded in the data set. Sample data is the smallest unit that constitutes a data set and is used to represent specific data content of different times and types; The changed sample data is recorded as valid data, and the recording times of two adjacent valid data in the A data sets are queried one by one through the timestamp, and the duration between the recording times of the two adjacent valid data is recorded as the change duration; After eliminating the maximum value of the change duration, the maximum value of the remaining change duration is recorded as the extreme value of the fluctuation duration; The digital features of two adjacent valid data are extracted respectively, and the absolute value is taken after the difference between the two digital features to obtain the change amplitude, and the number of change amplitudes greater than the calibrated change value is counted to obtain the fluctuation value. The digital features are used to represent the size of the specific digital part in the valid data and serve as the numerical basis for judging whether the valid data has changed. The calibrated change value is used to limit the minimum value of the change amplitude when participating in the fluctuation value statistics, which can ensure that the data change amplitude of the valid data in the fluctuation value is large.

[0020] After the interval parameters are collected, the interval duration needs to be calculated according to the interval parameters, and observation points are planned on the timeline based on the interval duration. The observation points at this time can be used as the time limit for collecting various data in the knitted fabric production line in the future. At the same time, the planned observation points can also mark the timeline for intervals. In this embodiment, the observation point is not a specific moment, but a time unit containing multiple continuous state time points; The marking process of observation points is as follows: The standard duration of the knitted fabric production line is pre-set, and the interval duration is calculated based on the standard duration, the extreme value of the fluctuation duration and the fluctuation value; the standard duration refers to the pre-set duration between two adjacent data collection moments in the knitted fabric production line under ideal conditions, so as to provide reasonable data support for the calculation of the interval duration of the actual observation point; The calculation formula for the interval duration is: ; In the formula, is the interval duration, is the standard duration, is the extreme value of the fluctuation duration, is the fluctuation value; Take an interval as the planning interval, take the data collection point as the first observation point, and plan B non-adjacent observation points on the timeline.

[0021] It should be noted that the duration between any two observation points is equal to ensure that the data of the knitted fabric production line at each observation point can maintain independence, thereby effectively avoiding the phenomenon of front-to-back adhesion of various data collected when the time is too close.

[0022] The equipment data collection module divides the knitted fabric production line into sections and collects the comprehensive equipment data of the knitted fabric production line at the observation point based on the comprehensive coverage criterion. The comprehensive equipment data includes the feed speed value, vibration drop value, regional tension and winding circle value; The section is used to specifically represent the different production stages in the overall production process of the knitted fabric production line, so as to distinguish the production and processing processes of the front-end feeding, middle-end processing and back-end winding of the knitted fabric production line; The sections include a feeding section, a knitting section and a winding section; specifically, the feeding section, the knitting section and the winding section are respectively used to represent the specific equipment included in the production and processing flow of front-end feeding, mid-end processing and back-end winding.

[0023] Comprehensive equipment data is used to comprehensively represent the diversified equipment data of the equipment in the overall production process of the knitted fabric production line. When collecting comprehensive equipment data, in order to maintain the data collection effect of the entire process of the knitted fabric production line, it is necessary to collect the comprehensive equipment data under the restriction of the comprehensive coverage criterion, so as to ensure that the comprehensive equipment data can comprehensively represent the equipment operation status of the feeding section, knitting section and winding section of the knitted fabric production line; The comprehensive coverage criterion is: the source sections of the comprehensive equipment data include the feeding section, knitting section and winding section; so that the comprehensive equipment data can come from different sections, realizing the data collection effect of multiple equipment, multiple sections and multiple dimensions in the knitted fabric production line.

[0024] Comprehensive equipment data includes feed speed value, vibration drop value, area tension and winding circle value; The feed speed value refers to the maximum value of the rotation speed of the feed motor in the feed section of the knitted fabric production line when it is at the observation point, which can be used to represent the working status of the feed motor in the feed section; the feed speed value is obtained by taking the maximum value after the speed sensor installed on the feed motor monitors the rotation speed of B observation points.

[0025] The vibration drop value refers to the vibration amplitude of the knitting head of the knitting section of the knitted fabric production line at the observation point, which can be used to represent the stability performance of the knitting head in the knitting section. The larger the vibration drop value, the worse the stability performance of the knitting head. The collection process of vibration drop value is as follows: All the moments within B observation points are marked one by one to obtain C observation moments, and the vibration value of the knitting machine head at the C observation moments is monitored in real time by a vibration sensor installed on the knitting machine head. After removing the maximum and minimum values ​​of the vibration value, the remaining C-2 vibration values ​​are recorded as available vibration values; After subtracting the maximum value of the available vibration value from the minimum value of the available vibration value, B vibration drop values ​​are obtained; The calculation formula for the vibration drop value is: ; In the formula, For the The vibration drop value of each observation point is =1,2,...,B, For the The maximum value of the available vibration value of the observation points is For the The minimum value of the available vibration values ​​at the observation points.

[0026] Regional tension refers to the tensioning support force of the tensioning mechanism on the knitted fabric in the knitting section of the knitted fabric production line at the observation point, which can be used to represent the tensioning performance of the knitted fabric in the knitting section. When collecting regional tension, the pressure values ​​at C observation times in B observation points are monitored by the pressure sensor installed on the tensioning mechanism, and all pressure values ​​are accumulated and averaged to obtain the value.

[0027] The winding circle value refers to the number of revolutions of the winding shaft in the winding section of the knitted fabric production line at the observation point, which can be used to indicate the winding stability performance of the winding shaft. The collection process of winding circle value is as follows: A camera installed on the winding machine frame is used to shoot a bird's-eye view video of the winding shaft in real time, and an image of the bird's-eye view video at the first observation time of B observation points is intercepted to obtain an initial image; The axis center and axis edge of the winding shaft are identified in the initial image by computer vision technology, and an edge point is randomly marked on the axis edge. The axis center and the edge point are connected by straight lines to obtain the center-edge line, and the initial position of the center-edge line in the initial image is identified; the axis center refers to the point where the center position of the winding shaft is located, and the axis edge refers to the outer edge of the roller of the winding shaft. By connecting the axis center and the axis edge by straight lines, the initial position of the winding shaft can be marked in the initial image, and accurate position limitation is provided for the subsequent identification and statistics of the number of rotations of the winding shaft; Identify the real-time position of the center edge line in the overhead video. When the real-time position of the center edge line coincides with the initial position once, increase the number of rotations of the reel by 1. From the first observation moment of B observation points to the last observation moment, the number of rotations of the reel in the overhead video is counted to obtain B first-circle values; Intercept the image of the last observation time of the overhead video at B observation points to obtain the termination image, and mark the real-time position of the center edge of the termination image; Along the rotation direction of the reel, measure the angle from the real-time position of the center edge of the end image to the initial position, record it as the rotation angle, and divide the rotation angle by the angle of one rotation to obtain B second circle values; The calculation formula for the second circle value is: ; In the formula, For the The second circle value of observation points, For the The rotation angle of the observation point, is the angle of one rotation; specifically, the angle of one rotation is 360 degrees; After sequentially adding the B first-circle values ​​to the B second-circle values, B winding-circle values ​​are obtained; The calculation formula for the winding circle value is: ; In the formula, For the The winding circle value of each observation point, For the The first circle value of each observation point.

[0028] The first model prediction module predicts the equipment status of the knitted fabric production line at the next observation point through the pre-trained first prediction model; The first prediction model is an artificial intelligence model used to analyze and predict the comprehensive equipment data of the knitted fabric production line, so that the comprehensive equipment data of the knitted fabric production line can be used as input data of the first prediction model to analyze and predict the equipment status corresponding to the comprehensive equipment data, and the equipment status is used to represent the operation status of the equipment of the knitted fabric production line at a specific future moment in a dimension; The equipment status includes normal status and abnormal status; the normal status and abnormal status correspond to the normal operation status and abnormal operation status of the equipment of the knitted fabric production line respectively. The equipment status is obtained by collecting a large number of historical feed speed values, vibration drop values, regional tension and winding circle values ​​corresponding to the specific operation of the equipment to determine whether it is faulty.

[0029] When analyzing and predicting the equipment operation status of the knitted fabric production line at a future time by using the first prediction model, it is necessary to train the first prediction model in advance; The training process of the first prediction model is as follows: Collect multiple sets of comprehensive equipment data of knitted fabric production lines in normal and abnormal states in advance; The comprehensive equipment data is converted into a plurality of first feature vectors using a sliding window method, the equipment state is converted into a first label corresponding to the comprehensive equipment data according to a first sliding step, the normal state and the abnormal state are converted into digital labels respectively, the normal state is converted into 0, and the abnormal state is converted into 1, one first feature vector corresponds to one first label, and constitutes a group of first training data, and multiple groups of first training data constitute a first training set; The comprehensive equipment data is arranged in the order of observation points, and the first predicted time step, the first sliding step and the first sliding window length are calibrated. The first eigenvector is used as the input of the first prediction model, and the equipment state of the next observation point after the first time step is predicted as the output. The subsequent equipment state of each first training set is used as the first prediction target. The first prediction model is trained with the minimized sum of the first prediction errors as the first training target to generate a first prediction model that predicts the equipment state of the next observation point based on the comprehensive equipment data.

[0030] The prediction process of normal state and abnormal state is as follows: When the output of the first prediction model is 0, it means that the equipment of the knitted fabric production line at the next observation point maintains normal operation, and the predicted next observation point is in a normal state; When the output of the first prediction model is 1, it indicates that the equipment of the knitted fabric production line at the next observation point does not maintain normal operation, and the next observation point is predicted to be in an abnormal state.

[0031] The technical solutions in the above-mentioned equipment data acquisition module and the first model prediction module are applied in an actual knitted fabric production line, and B observation points are planned in chronological order, and the B observation points are numbered in ascending order, and the numbers are recorded as 1, 2, 3, ..., B respectively. At the same time, the comprehensive equipment data and equipment status of the B observation points are collected, and the obtained data are summarized and analyzed to form a first prediction data table, as shown in Table 1 below: Table 1: First forecast data table

[0032] The quality data collection module collects the comprehensive quality data of the knitted fabric production line at the observation point. The comprehensive quality data includes unit density value, local flatness and boundary width value; Comprehensive quality data is used to express the quality of knitted fabrics in the knitted fabric production line in a variety of data, which can be used to comprehensively display the production quality performance of knitted fabrics in the knitted fabric production line; Comprehensive quality data include unit density values, local flatness and boundary width values; The unit density value refers to the compactness of the knitted fabric in the unit area of ​​the knitted fabric production line, which can be used to express the density performance of the knitted product; The process of collecting unit density values ​​is as follows: The observation time at the middle position of the B observation points is recorded as the target time, a bird's-eye view image of the knitted fabric at the target time is captured by a camera, the bird's-eye view image is converted into a grayscale image, all the pixels in the B grayscale images are marked one by one, and the pixel values ​​of all the pixels are identified; Mark the four corner points of the grayscale image, connect the four corner points diagonally to form two diagonal lines, measure the lengths of the two diagonal lines respectively, and record one seventh of the minimum length value as the target length; Take the intersection of the two diagonals as the center of the circle and the target length as the radius to draw a central circle. Draw an outer circle with the same area as the central circle in a ring structure outside the central circle, and then sum up the central circle and the outer circle to obtain pixel circles; by drawing a central circle at the intersection of the diagonals, the central area can be located in the grayscale image, and by distributing the outer circles in a ring outside the central circle, it is possible to achieve multi-position and uniform position recognition and data collection effects in the grayscale image, avoiding the limitations of data collection at a single position, and at the same time improve the accuracy of the data, avoid the interference caused by the negative effects of ambient light and other negative effects at the edge of the grayscale image, and effectively improve the calculation accuracy of the unit density value in the grayscale image; The pixels whose pixel values ​​are lower than the calibration pixel threshold are recorded as target pixels, and the pixels are counted one by one. The number of target pixels within the pixel circle is obtained The pixel threshold calibration is used to represent the pixel value of the specific content represented by the pixel in the grayscale image, so that the pixel points greater than the pixel threshold calibration correspond to the knitted fabric, and the pixel points less than the pixel threshold calibration correspond to the gap between the knitted fabrics, thereby providing an accurate numerical basis for the identification of the target pixel points. Will After dividing the target value by the total number of pixels in the pixel circle, we get The density of the sub- After accumulating the density of each particle, we find the average value and obtain B unit density values; The unit density value is calculated as: ; In the formula, For the The unit density value of the observation point, For the The first observation point The target value of the pixel circle, For the The first observation point The total number of pixels in a pixel circle.

[0033] Local flatness refers to the degree of flatness and smoothness of the knitted fabric in the local position of the knitted fabric production line, which can be used to indicate the flatness of the knitted product; The process of collecting local flatness is as follows: At the target moment of B observation points, the side view image of the knitted fabric is captured by a camera, and the location of the knitted fabric is identified by computer vision technology to obtain the knitted area; After drawing lines at the positions of the upper and lower edges of the knitted area in the horizontal direction, draw parallel upper and lower boundary lines; The design thickness of the knitted fabric is found out, the lower boundary line is kept stationary, and the distance from the upper boundary line to the lower boundary line is adjusted in the vertical direction until the distance from the upper boundary line to the lower boundary line is consistent with the design thickness, and then the adjustment is stopped; the upper boundary line is adjusted by keeping the lower boundary line stationary, so that the distance between the upper boundary line and the lower boundary line is consistent with the design thickness, so that the upper boundary line and the lower boundary line can be consistent with the standard thickness of the knitted fabric, and at the same time, an accurate position definition basis can be provided for whether the surface fabric of the knitted fabric is a bulge in actual processing of the knitted fabric; The number of pixels located above the upper boundary line and the number of pixels located between the upper boundary line and the lower boundary line are counted respectively to obtain a convexity value and a flatness value, and after dividing the convexity value by the flatness value, B local flatness values ​​are obtained; The calculation formula for local flatness is: ; In the formula, For the The local flatness of each observation point, For the The convex value of the observation point, For the The smoothed value of each observation point.

[0034] The boundary width value refers to the width between the two sides of the knitted fabric of the knitted fabric production line, which can represent the width performance of the knitted product; the boundary width value is obtained by measuring the length between the two sides of the knitted fabric in B grayscale images.

[0035] The second model prediction module predicts the quality status of the knitted fabric production line at the next observation point through the pre-trained second prediction model; The second prediction model is an artificial intelligence model used to analyze and predict the comprehensive quality data of the knitted fabric production line, so that the comprehensive quality data of the knitted fabric production line can be used as input data of the second prediction model to analyze and predict the fabric state corresponding to the comprehensive quality data, and the quality state is used to represent the operation status of the knitted fabric of the knitted fabric production line at a specific future moment in another dimension; The quality status includes a high quality status and a low quality status; the high quality status and the low quality status respectively correspond to the high quality and low quality of the knitted fabric in the knitted fabric production line, and the quality status is obtained by collecting a large amount of historical unit density values, local flatness and boundary width values ​​corresponding to the specific quality data of the knitted fabric.

[0036] When analyzing and predicting the quality of knitted fabrics of the knitted fabric production line at a future time by using the second prediction model, it is necessary to train the second prediction model in advance; The training process of the second prediction model is as follows: Collect multiple sets of comprehensive quality data of knitted fabric production lines in high-quality and low-quality states in advance; The comprehensive quality data is converted into a plurality of second feature vectors using a sliding window method, the quality state is converted into a second label corresponding to the comprehensive quality data according to a second sliding step, the high-quality state and the low-quality state are converted into digital labels respectively, the high-quality state is converted into 2, and the low-quality state is converted into 3, one second feature vector corresponds to one second label, and constitutes a group of second training data, and multiple groups of second training data constitute a second training set; The comprehensive quality data are arranged in the order of the observation points, and the predicted second time step, the second sliding step and the second sliding window length are calibrated. The second eigenvector is used as the input of the second prediction model, and the quality state of the next observation point after the second time step is predicted as the output. The subsequent quality state of each second training set is used as the second prediction target. The second prediction model is trained with the minimized sum of the second prediction errors as the second training target to generate a second prediction model that predicts the quality state of the next observation point based on the comprehensive quality data.

[0037] It should be noted that the first prediction model and the second prediction model adopt any one of the CNN neural network model or AlexNet.

[0038] The prediction process of high-quality state and low-quality state is as follows: When the output of the second prediction model is 2, it means that the knitted fabric of the knitted fabric production line at the next observation point has a higher quality, and the predicted next observation point is in a high-quality state; When the output of the second prediction model is 3, it means that the knitted fabric of the knitted fabric production line at the next observation point does not have a high quality, and the next observation point is predicted to be in a low quality state.

[0039] The technical solutions in the above-mentioned quality data collection module and the second model prediction module are applied in the actual knitted fabric production line, and the comprehensive quality data and quality status of B observation points are collected at the same time, and the obtained data are summarized and analyzed to form a second prediction data table, as shown in Table 2 below: Table 2: Second prediction data table

[0040] The production and processing control module combines and analyzes the equipment status and quality status at the same observation point, determines the production and processing status of the knitted fabric production line, and controls the knitted fabric production line to automatically process the knitted fabric; After predicting the equipment status and quality status of a certain observation point in the future, the comprehensive status of the knitted fabric production line can be expressed according to the predicted combination results of the equipment and fabric of the knitted fabric production line, and the comprehensive status is recorded as the production and processing status; The production and processing status includes the shutdown status and the maintenance status; the shutdown status and the maintenance status are used to respectively indicate the poor and good comprehensive status of the knitted fabric production line; The process of determining the shutdown state and the maintenance state is as follows: Combine and analyze the predicted equipment status and quality status of the same observation point in the future; When the equipment status and quality status of the same observation point in the future are normal and high-quality respectively, the equipment of the knitted fabric production line at the same observation point in the future is normal and the quality of the knitted fabric is high, and the production and processing status is determined to be the maintenance status; When the equipment status and quality status at the same observation point in the future are not abnormal and low-quality respectively, at this time, the equipment at the same observation point of the knitted fabric production line in the future is abnormal and the quality of the knitted fabric is low, and the production and processing status is determined to be a shutdown state.

[0041] After determining the specific production and processing status of the knitted fabric production line, it is necessary to formulate specific measures for optimizing the control of the knitted fabric production line according to the specific production and processing status, so that the optimized control measures can serve as the basis for controlling the shutdown and maintenance and continuous operation of the knitted fabric production line; The control process of the knitted fabric production line is as follows: When the production and processing state of the knitted fabric production line at the future Eth observation point is the maintenance state, there is no need to perform optimization control operations on the relevant parameters of the knitted fabric production line. The control terminal controls the knitted fabric production line to continue processing the knitted fabric at the Eth observation point. When the production and processing status of the knitted fabric production line at the Eth observation point in the future is a shutdown state, it is necessary to optimize the control operation of the relevant parameters of the knitted fabric production line. The control terminal controls the knitted fabric production line to stop processing knitted fabric at the E-1th observation point.

[0042] It should be noted that when the knitted fabric production line is controlled to stop production and processing at the E-1 observation point, it is necessary to further analyze the comprehensive equipment data and comprehensive quality data to identify the specific data with abnormal phenomena at the E-1 observation point; Specifically, the feed speed value, vibration drop value, regional tension, winding circle value, unit density value, local flatness and boundary width value of the E-1th observation point are compared with the preset speed safety value, vibration safety value, tension safety degree, winding safety value, density safety value, flatness safety degree and width safety value respectively, and the data greater than or less than the corresponding safety value or safety degree is regarded as abnormal data indicating abnormal phenomenon. Then, according to the specific type of abnormal data, the control server sends the corresponding control instruction to the control terminal, and optimizes the parameters of the specific equipment corresponding to the abnormal data at the E-1th observation point. The optimization and adjustment operations include but are not limited to increasing the value of the abnormal data and decreasing the value of the abnormal data, thereby ensuring that the predicted production and processing state of the knitted fabric production line at the Eth observation point is converted from a shutdown state to a maintenance state, and finally achieving the automated control processing effect of the knitted fabric production line, ensuring that the knitted fabric production line can always be in a processing state with normal equipment and high-quality knitted fabrics.

[0043] In this embodiment, by setting observation points distributed at intervals, it is possible to provide a time limit for the collection of good and bad evaluation data of the actual processing status of the knitted fabric production line, ensure that the relevant data collected subsequently can comprehensively represent the different time points of the knitted fabric production line, and ensure that the collected data are independent. At the same time, by separately collecting the equipment data and the knitted fabric data in the knitted fabric production line, it is possible to represent the processing status of the knitted fabric production line from the two aspects of equipment and knitted fabric, and by combining two independently running prediction models, it is possible to make an advance prediction of the actual status of the knitted fabric production line equipment and knitted fabric at future observation points, and make a comprehensive and accurate prediction of the comprehensive processing status of the knitted fabric production line through the actual status of the equipment and knitted fabric at future observation points, effectively avoiding the limitations and low accuracy of the single-dimensional data source and the single-function prediction model in predicting the future processing status, thereby achieving the advance prediction effect of the processing status of the knitted fabric production line at a future moment, and then being able to provide effective solution guidance for the problems existing in the knitted fabric production line in advance and accurately before the knitted fabric production line is about to have abnormal failure phenomena, and ultimately achieve the automation control effect of the knitted fabric production line.

[0044] Example 2: Please refer to Figure 2 As shown, the part not described in detail in this embodiment is described in the first embodiment, and a production line automation control method based on knitted fabric processing is provided, which is applied to a control server and is implemented based on a production line automation control system based on knitted fabric processing, including: S1: After the data collection point, collect the extreme value of the fluctuation duration and the fluctuation value of the knitted fabric production line, calculate the interval duration, and plan the observation points of the interval distribution; S2: Divide the knitted fabric production line into the feeding section, knitting section and winding section, and collect the comprehensive equipment data of the knitted fabric production line at the observation point based on the full coverage criterion; S3: predicting the equipment status of the knitted fabric production line at the next observation point through the pre-trained first prediction model; S4: Perform visual image analysis on knitted fabrics and collect comprehensive quality data of knitted fabric production lines at observation points; S5: predicting the quality status of the knitted fabric production line at the next observation point through the pre-trained second prediction model; S6: Combine and analyze the equipment status and quality status of the same observation point to determine the production and processing status of the knitted fabric production line at the future observation point, and control the automated processing of the knitted fabric production line.

[0045] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A production line automation control system based on knitted fabric processing, applied to a control server, characterized in that: include: The observation point planning module is used to collect the extreme value of the fluctuation duration and the fluctuation value of the knitted fabric production line after the data collection point, calculate the interval duration, and plan the observation points of the interval distribution; The equipment data acquisition module is used to divide the knitted fabric production line into a feeding section, a knitting section and a winding section. Based on the full coverage principle, the comprehensive equipment data of the knitted fabric production line at the observation point is collected. The comprehensive equipment data includes the feeding speed value, the vibration drop value, the regional tension and the winding circle value; The comprehensive coverage criteria are: the source sections of the comprehensive equipment data include the feeding section, knitting section and winding section; A first model prediction module is used to predict the equipment state of the knitted fabric production line at the next observation point by using a pre-trained first prediction model; The quality data acquisition module is used to perform visual image analysis on knitted fabrics and collect comprehensive quality data of the knitted fabric production line at the observation point. The comprehensive quality data includes unit density value, local flatness and boundary width value; A second model prediction module is used to predict the quality status of the knitted fabric production line at the next observation point by using a pre-trained second prediction model; The production and processing control module is used to combine and analyze the equipment status and quality status of the same observation point, determine the production and processing status of the knitted fabric production line at the future observation point, and control the automated processing of the knitted fabric production line.

2. The automatic control system for a production line based on knitted fabric processing according to claim 1, characterized in that: The collection process of the extreme value of the fluctuation duration and the fluctuation value is as follows: The period between the data collection point and the current time is recorded as the observation period, and the A data sets recorded in the observation period are queried through the database, and the sample data in the A data sets are arranged in sequence according to the order of recording; The changed sample data is recorded as valid data, and the recording times of two adjacent valid data in the A data sets are queried one by one through the timestamp, and the duration between the recording times of the two adjacent valid data is recorded as the change duration; After eliminating the maximum value of the change duration, the maximum value of the remaining change duration is recorded as the extreme value of the fluctuation duration; The digital features of two adjacent valid data are extracted respectively, the difference between the two digital features is taken and the absolute value is obtained to obtain the change amplitude, and the number of change amplitudes greater than the calibrated change value is counted to obtain the fluctuation value.

3. The automatic control system for a production line based on knitted fabric processing according to claim 2 is characterized in that: The marking process of the observation points is as follows: The standard duration of the knitted fabric production line is pre-set, and the interval duration is calculated based on the standard duration, the extreme value of the fluctuation duration and the fluctuation value; The calculation formula for the interval duration is: ； In the formula, is the interval duration, is the standard duration, is the extreme value of the fluctuation duration, is the fluctuation value; Take an interval as the planning interval, take the data collection point as the first observation point, and plan B non-adjacent observation points on the timeline.

4. The production line automation control system based on knitted fabric processing according to claim 3 is characterized in that: The collection process of vibration drop value is as follows: All the moments within B observation points are marked one by one to obtain C observation moments, and the vibration value of the knitting machine head at the C observation moments is monitored in real time by a vibration sensor installed on the knitting machine head. After removing the maximum and minimum values ​​of the vibration value, the remaining C-2 vibration values ​​are recorded as available vibration values; After subtracting the maximum available vibration value from the minimum available vibration value, B vibration drop values ​​are obtained.

5. The production line automation control system based on knitted fabric processing according to claim 4 is characterized in that: The collection process of winding circle value is as follows: A camera installed on the winding machine frame is used to shoot a bird's-eye view video of the winding shaft in real time, and an image of the bird's-eye view video at the first observation time of B observation points is intercepted to obtain an initial image; The axis center and axis edge of the reel are identified in the initial image by computer vision technology, a side point is randomly marked on the side edge, the axis center and the side point are connected by a straight line to obtain the center-edge line, and the initial position of the center-edge line in the initial image is identified; Identify the real-time position of the center edge line in the overhead video. When the real-time position of the center edge line coincides with the initial position once, increase the number of rotations of the reel by 1. From the first observation moment of B observation points to the last observation moment, the number of rotations of the reel in the overhead video is counted to obtain B first-circle values; Intercept the image of the last observation time of the overhead video at B observation points to obtain the termination image, and mark the real-time position of the center edge of the termination image; Along the rotation direction of the reel, measure the angle from the real-time position of the center edge of the end image to the initial position, record it as the rotation angle, and divide the rotation angle by the angle of one rotation to obtain B second circle values; After sequentially adding the B first-circle values ​​to the B second-circle values, B winding-circle values ​​are obtained.

6. The automatic control system for a production line based on knitted fabric processing according to claim 5, characterized in that: The equipment status includes normal status and abnormal status; The prediction process of normal state and abnormal state is as follows: When the output of the first prediction model is 0, the predicted equipment state of the next observation point of the knitted fabric production line is normal; When the output of the first prediction model is 1, the predicted equipment state of the next observation point of the knitted fabric production line is an abnormal state.

7. The automatic control system for a production line based on knitted fabric processing according to claim 6, characterized in that: The process of collecting unit density values ​​is as follows: The observation time at the middle position of the B observation points is recorded as the target time, a bird's-eye view image of the knitted fabric at the target time is captured by a camera, the bird's-eye view image is converted into a grayscale image, all the pixels in the B grayscale images are marked one by one, and the pixel values ​​of all the pixels are identified; Mark the four corner points of the grayscale image, connect the four corner points diagonally to form two diagonal lines, measure the lengths of the two diagonal lines respectively, and record one seventh of the minimum length value as the target length; Take the intersection of the two diagonals as the center of the circle and the target length as the radius to draw a central circle. Draw an outer circle with the same area as the central circle in a ring structure outside the central circle, and then sum up the central circle and the outer circle to obtain pixel circle; The pixels whose pixel values ​​are lower than the calibration pixel threshold are recorded as target pixels, and the pixels are counted one by one. The number of target pixels within the pixel circle is obtained target value; Will After dividing the target value by the total number of pixels in the pixel circle, we get The density of the sub- The density of each sub-unit is accumulated and averaged to obtain B unit density values.

8. The automatic control system for a production line based on knitted fabric processing according to claim 7, characterized in that: The process of collecting local flatness is as follows: At the target moment of B observation points, the side view image of the knitted fabric is captured by a camera, and the location of the knitted fabric is identified by computer vision technology to obtain the knitted area; After drawing lines at the positions of the upper and lower edges of the knitted area in the horizontal direction, draw parallel upper and lower boundary lines; Find out the design thickness of the knitted fabric, keep the lower boundary line still, and adjust the distance from the upper boundary line to the lower boundary line in the vertical direction until the distance from the upper boundary line to the lower boundary line is consistent with the design thickness, and then stop adjusting; The number of pixels above the upper boundary line and the number of pixels between the upper boundary line and the lower boundary line are counted respectively to obtain a convexity value and a flatness value. After dividing the convexity value and the flatness value, B local flatness values ​​are obtained.

9. The production line automation control system based on knitted fabric processing according to claim 8, characterized in that: The quality state includes a high quality state and a low quality state; The prediction process of high-quality state and low-quality state is as follows: When the output of the second prediction model is 2, the quality state of the next observation point of the knitted fabric production line is predicted to be a high quality state; When the output of the second prediction model is 3, the predicted quality state of the next observation point of the knitted fabric production line is a low quality state.

10. The production line automation control system based on knitted fabric processing according to claim 9, characterized in that: The production and processing status includes the shutdown status and the maintenance status. The determination process of the shutdown status and the maintenance status is as follows: Combine and analyze the predicted equipment status and quality status of the same observation point in the future; When the equipment status and quality status of the same observation point in the future are respectively normal and high-quality, the production and processing status is determined to be the maintenance status; When the equipment status and quality status of the same observation point in the future are not abnormal and low-quality respectively, the production and processing status is determined as the shutdown status; The control process of the knitted fabric production line is as follows: When the production and processing state of the knitted fabric production line at the future Eth observation point is the maintenance state, the control terminal controls the knitted fabric production line to continue processing the knitted fabric at the Eth observation point; When the production and processing state of the knitted fabric production line at the Eth observation point in the future is a shutdown state, the control terminal controls the knitted fabric production line to stop processing the knitted fabric at the E-1th observation point.

Images