Method and system for monitoring production quality of home textile fabric based on industrial data

Abstract

The invention relates to the technical field of home textile fabric production quality monitoring, in particular to a home textile fabric production quality monitoring method and system based on industrial data. Comprising the following steps: firstly, selecting a home textile fabric template from a template library by acquiring planned finished product information, and calculating the difference degree between the home textile fabric template and a planned finished product; according to the difference degree, initial equipment parameters are determined, and the equipment is set through the parameters. Next, in the production process, the home textile fabric quality monitoring model is adopted to carry out image acquisition, processing and analysis on the target home textile fabric, and a bottom layer and a top layer in the image are identified and separated; the quality monitoring comprises the steps of analyzing the characteristics of the bottom-layer and top-layer images, judging the qualification degree and generating a qualification degree vector; and if the monitoring result has a quality problem, adjusting equipment parameters, and continuously performing quality monitoring. And finally, equipment parameters are updated according to the unqualified data through the deduction model, the production quality is ensured, and the number of unqualified products is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of home textile fabric production quality monitoring, and in particular to a home textile fabric production quality monitoring method and system based on industrial data. Background Art

[0002] The production of home textile fabrics is a highly complex system engineering involving multiple links, usually including spinning, weaving, dyeing and finishing and other complex processes. In these links, various parameters such as temperature and humidity, equipment operation status and raw material quality will significantly affect the quality of the final fabric. Therefore, strict quality control must be implemented during the production process to ensure that each link can strictly meet the predetermined quality standards.

[0003] However, traditional quality control methods mainly focus on testing finished products, which has certain limitations. For example, existing methods are difficult to detect potential problems in the production process in a timely manner, and usually rely on the later finished product testing to assess quality. Although this method helps to identify defects, it cannot detect problems in the early stages of production, and thus cannot effectively guarantee the overall quality of home textile fabrics, resulting in an increase in the rate of defective products.

[0004] In order to solve this problem, the present invention proposes a method and system for monitoring the production quality of home textile fabrics based on industrial data, aiming to improve the quality control capability in the production process and the qualified rate of products. Summary of the invention

[0005] The purpose of the present invention is to provide a home textile fabric production quality monitoring method and system based on industrial data, which are used to monitor the quality of home textile fabrics during production. The present invention mainly starts from the following points: first, by obtaining planned finished product information and automatically adjusting equipment parameters according to their differences, the initial equipment settings are ensured to be accurate, thereby improving production efficiency; second, a home textile fabric quality monitoring model is used to perform real-time image processing and quality assessment, including separation and feature extraction of the bottom layer and the top layer, to ensure quality control of each production link; third, through multi-level quality monitoring sublayers (bottom layer, top layer and joint quality monitoring), the fabric is carefully inspected, unqualified items are discovered and marked in time, and the quality consistency of the final product is guaranteed; finally, when problems are found in the monitoring results, the system can intelligently adjust the parameters of the production equipment and perform continuous quality monitoring to achieve comprehensive optimization and quality improvement of the home textile fabric production process.

[0006] To achieve the above object, the present invention provides the following technical solutions: A method for monitoring the production quality of home textile fabrics based on industrial data, comprising: Obtain the planned finished product information of the home textile fabrics to be processed; Further, initial equipment parameters are obtained according to the planned finished product information; wherein the process of obtaining the initial equipment parameters includes: Selecting a home textile fabric template from a template library according to the planned finished product information; Further, calculating the difference between the planned finished product information and the home textile fabric template; When the difference is 0, the standard equipment parameters of the home textile fabric template are used as the initial equipment parameters of the home textile fabric to be processed; When the difference is not 0, calculating the deviation value between the home textile fabric to be processed and the home textile fabric template, and constructing a home textile fabric deviation matrix; According to the deviation values ​​in the home textile fabric deviation matrix, the equipment attributes of each production step are obtained; Further, establishing a device-parameter mapping between the device attributes of the home textile fabric template and the standard device parameters; Furthermore, the standard equipment correction parameters are obtained through the equipment parameter correction function; wherein the equipment parameter correction function is: ;in, Corrected parameters for standard equipment; is the standard equipment parameter; A mapping matrix corresponding to the device-parameter mapping; is the home textile fabric deviation matrix; is a nonlinear correction function; For production environment data; Furthermore, the standard equipment correction parameters are used as initial equipment parameters.

[0007] Further, the device is set using the initial device parameters; Further, the home textile fabric to be processed is produced as a target object; Further, the quality monitoring of the target object is performed using a home textile fabric quality monitoring model, including: a collection layer for collecting images of the target object during the monitoring process; A preprocessing layer is used to perform image processing on the target object image to obtain a standard target object image; wherein the preprocessing layer includes: performing background segmentation on the target object image to obtain a first target object image; performing denoising on the first target object image to obtain a second target object image; performing smoothing on the second target object image to obtain a third target object image; performing edge detection on the third target object image to obtain a fourth target object image; and performing image enhancement on the fourth target object image to obtain the standard target object image.

[0008] Identify and separate layers, for identifying and separating the bottom layer and the top layer of the standard target object image, The specific process includes: starting from the upper left corner of the standard target object image, the recognition frame is moved horizontally to obtain the recognition frame image and the recognition frame coordinates; wherein the recognition frame coordinates take the center point of the standard target object image as the origin, and the coordinates are marked as (0,0); Further, extracting features of the recognition frame image to obtain a first feature vector; Further, the first feature vector is input into a classification model to obtain a classification result; wherein the classification result includes two classification types: bottom layer and top layer; Further, the standard target object image is cropped according to the classification result to obtain a bottom layer image and a top layer image; Quality monitoring layer: used to monitor the quality of the bottom layer image and the top layer image; the quality monitoring layer includes: a bottom layer quality monitoring sublayer, a top layer quality monitoring sublayer and a joint quality monitoring sublayer; wherein the specific process of the quality monitoring layer includes: Input the bottom layer image into the bottom layer quality monitoring sublayer to obtain the bottom layer features of the target object; wherein the bottom layer features of the target object include: bottom layer specification features, bottom layer structure features, bottom layer color features and bottom layer surface features; Perform bottom quality analysis on the bottom features of the target object to obtain a bottom fabric qualification vector; wherein the bottom fabric qualification vector includes: bottom fabric specification qualification, bottom fabric structure qualification, bottom fabric color qualification and bottom fabric defect qualification; when any qualification in the bottom fabric qualification vector does not meet the bottom fabric standard, mark the bottom fabric unqualified type and obtain the corresponding unqualified bottom fabric data; Input the top layer image into the top layer quality monitoring sublayer to obtain the top layer features of the target object; wherein the top layer features of the target object include: top layer shape features, top layer color features, top layer texture features and top layer integrity features; perform top layer quality analysis on the top layer features of the target object to obtain the top layer qualification vector of the fabric; wherein the top layer qualification vector of the fabric includes: top layer shape qualification, top layer color qualification, top layer texture qualification and top layer integrity qualification; when any qualification in the top layer qualification vector of the fabric does not meet the top layer standard, mark the top layer unqualified type and obtain the corresponding unqualified top layer data; Input the bottom layer image and the top layer image into the joint quality monitoring sublayer, and perform image superposition according to the standard target object image to obtain a target object joint image; extract joint features of the target object joint image; wherein the joint features include: position features, size features, and scale features; The expected error between the joint feature and the current production expectation is calculated to construct a production error vector; when the comprehensive error value of the production error vector exceeds the error tolerance, the joint unqualified type is marked, and the corresponding unqualified joint data is obtained.

[0009] An output layer, used to output the monitoring results of the quality monitoring layer; Further, when the monitoring result has no quality problem, the initial equipment parameters of the current production step are used; when the monitoring result has quality problem, the initial equipment parameters of the current production step are adjusted as equipment adjustment parameters, and quality monitoring is continuously performed; the specific process includes: Obtain the unqualified type and corresponding unqualified data of the target object; Further, the unqualified type and the unqualified data are characterized to obtain an unqualified feature vector; Further, the unqualified feature vector and the initial equipment parameters are input into a random forest model to obtain an equipment parameter influence vector; Further, a correlation analysis is performed on the device parameter influence vector to obtain a device parameter correlation matrix; Further, the current production expectation is obtained, and the current production expectation feature vector is obtained by characterization; Furthermore, the unqualified feature vector, the initial equipment parameters and the equipment parameter association matrix are input into the deduction model; when the error between the deduced unqualified feature vector and the current production expected feature vector is minimized, the initial equipment parameters are updated and the equipment adjustment parameters of the current production step are obtained.

[0010] A home textile fabric production quality monitoring system based on industrial data, the system comprising: A home textile fabric production plan generation unit, used to generate a production plan for home textile fabrics to be processed; Template unit, used to store templates of various home textile fabrics, as well as corresponding template information and standard equipment parameters; The initial equipment parameter generation unit is used to generate the initial parameters of each equipment in the production process of the home textile fabric to be processed; wherein, the initial equipment parameter generation unit: Selecting a home textile fabric template from a template library according to the planned finished product information; calculating the difference between the planned finished product information and the home textile fabric template; When the difference is 0, the standard equipment parameters of the home textile fabric template are used as the initial equipment parameters of the home textile fabric to be processed; When the difference is not 0, calculating the deviation value between the home textile fabric to be processed and the home textile fabric template, and constructing a home textile fabric deviation matrix; Further, according to the deviation values ​​in the home textile fabric deviation matrix, the equipment attributes of each production step are obtained; Further, establishing a device-parameter mapping between the device attributes of the home textile fabric template and the standard device parameters; Furthermore, the standard equipment correction parameters are obtained through the equipment parameter correction function; wherein the equipment parameter correction function is: ;in, Corrected parameters for standard equipment; is the standard equipment parameter; A mapping matrix corresponding to the device-parameter mapping; is the home textile fabric deviation matrix; is a nonlinear correction function; is production environment data; the standard equipment correction parameters are used as initial equipment parameters; Production unit, used for the production of home textile fabrics to be processed; The home textile fabric quality monitoring unit is used to monitor the quality of the home textile fabric to be processed during the production process; wherein the home textile fabric quality monitoring unit adopts a home textile fabric quality monitoring model to perform quality monitoring, including: Further, the home textile fabrics to be processed in the production unit are taken as target objects; Further, the target object image is acquired using the acquisition layer; Furthermore, the target object image is input into a preprocessing layer to obtain a standard target object image; wherein the preprocessing layer includes: performing background segmentation on the target object image to obtain a first target object image; performing denoising on the first target object image to obtain a second target object image; performing smoothing on the second target object image to obtain a third target object image; performing edge detection on the third target object image to obtain a fourth target object image; and performing image enhancement on the fourth target object image to obtain the standard target object image.

[0011] Furthermore, the standard target object image is input into the recognition separation layer to obtain a bottom layer image and a top layer image; wherein the recognition separation layer includes: starting from the upper left corner of the standard target object image through a recognition frame, moving horizontally to obtain a recognition frame image and recognition frame coordinates; extracting features of the recognition frame image to obtain a first feature vector; inputting the first feature vector into a classification model to obtain a classification result; and cropping the standard target object image according to the classification result to obtain the bottom layer image and the top layer image.

[0012] Furthermore, the bottom layer image and the top layer image are input into a quality monitoring layer to obtain a quality monitoring result; wherein the specific process of the quality monitoring layer includes: Input the bottom layer image into the bottom layer quality monitoring sublayer to obtain the bottom layer features of the target object; wherein the bottom layer features of the target object include: bottom layer specification features, bottom layer structure features, bottom layer color features and bottom layer surface features; Perform bottom quality analysis on the bottom features of the target object to obtain a bottom fabric qualification vector; wherein the bottom fabric qualification vector includes: bottom fabric specification qualification, bottom fabric structure qualification, bottom fabric color qualification and bottom fabric defect qualification; when any qualification in the bottom fabric qualification vector does not meet the bottom fabric standard, mark the bottom fabric unqualified type and obtain the corresponding unqualified bottom fabric data; Input the top layer image to the top layer quality monitoring sublayer to obtain the top layer features of the target object; wherein the top layer features of the target object include: top layer shape features, top layer color features, top layer texture features and top layer integrity features; Performing a top-level quality analysis on the top-level features of the target object to obtain a top-level quality vector of the fabric; wherein the top-level quality vector of the fabric includes: top-level shape quality, top-level color quality, top-level texture quality and top-level integrity quality; when any quality item in the top-level quality vector of the fabric does not meet the top-level standard, marking the top-level unqualified type and obtaining the corresponding unqualified top-level data; Inputting the bottom layer image and the top layer image into the joint quality monitoring sublayer, and performing image superposition according to the standard target object image to obtain a target object joint image; Extracting joint features of the joint image of the target object; wherein the joint features include: position features, size features and scale features; The expected error between the joint feature and the current production expectation is calculated to construct a production error vector; when the comprehensive error value of the production error vector exceeds the error tolerance, the joint unqualified type is marked, and the corresponding unqualified joint data is obtained.

[0013] Furthermore, the quality monitoring result is outputted using an output layer.

[0014] Equipment parameter optimization unit, used to optimize the initial parameters of the equipment; The feedback unit is used to provide real-time feedback on the quality monitoring results of the home textile fabrics to be processed and the equipment parameter adjustment status.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. This method selects a suitable home textile fabric template from the template library, calculates the difference and deviation matrix according to the planned finished product information, and combines the equipment parameter mapping and correction function to accurately adjust the production equipment, thereby achieving precise control of equipment parameters, optimizing the production process, and ensuring the production quality of home textile fabrics. When the difference is 0, the standard equipment parameters of the home textile fabric template are directly used; when the difference is not 0, the deviation value is calculated and the home textile fabric deviation matrix is ​​constructed. The equipment parameters are further adjusted through equipment-parameter mapping and correction function to ensure that the equipment can adapt to the production needs of different fabrics. In addition, production environment data (such as temperature, humidity, etc.) as part of the correction function can further improve the accuracy of parameter adjustment, ensure the stability and efficiency of the equipment in a changing environment, thereby effectively improving the quality and pass rate of home textile fabrics.

[0016] 2. The present invention proposes a home textile fabric quality monitoring model for monitoring the quality of home textile fabrics during production. The model generates standardized target images through multi-level image processing and feature extraction; classifies images by identifying separation layers, extracts bottom and top features, and combines the bottom, top and joint quality monitoring sublayers to comprehensively analyze key indicators such as fabric specifications, structures, colors, surface features, shapes, textures and integrity. Finally, the quality qualification vector and production errors of the fabric are output according to the monitoring results, unqualified items are identified in time, and their types are marked, thereby realizing accurate quality monitoring of home textile fabrics, providing strong analytical support for quality control in subsequent production processes and the production of high-quality home textile fabrics.

[0017] 3. The present invention proposes a method for adjusting equipment parameters in the production process of home textile fabrics according to the quality monitoring results. The method first obtains the unqualified type and data of the target object through a series of precise analysis steps, characterizes them, and generates unqualified feature vectors; then, the unqualified feature vectors and initial equipment parameters are input into the random forest model to obtain the equipment parameter influence vector; then, correlation analysis is performed ， Generate an equipment parameter association matrix; then characterize the current production expectations and input them into the deduction model as the optimization target to update the initial equipment parameters, thereby significantly improving the product quality of home textile fabrics and reducing the defective rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A flowchart of a method for monitoring the production quality of home textile fabrics based on industrial data provided by an embodiment of the present invention; Figure 2 A structural diagram of a home textile fabric production quality monitoring system based on industrial data provided by an embodiment of the present invention; Figure 3A flowchart of obtaining initial device parameters provided by an embodiment of the present invention; Figure 4 A schematic diagram of acquiring an image using a recognition frame provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0019] 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.

[0020] Home textile fabrics are fabrics specifically used to make home textiles such as bedding, curtains, sofa covers, carpets, etc. As consumers' requirements for the comfort, functionality and aesthetics of their home environments continue to increase, the demand for home textile fabrics and their quality standards are also increasing.

[0021] As people's requirements for home textile fabric production process and quality control become increasingly stringent, quality control has become a crucial link, which directly affects the performance and reliability of the product. Therefore, in the production process of home textile fabrics, every link must be strictly controlled. However, traditional quality monitoring methods mainly focus on finished product testing, which has certain limitations. Existing methods are difficult to detect potential problems in a timely manner, and usually can only rely on post-testing to judge quality. Although this method can identify defects, it cannot detect problems in advance during the production process, resulting in an increase in the rate of defective products.

[0022] In order to solve the shortcomings of the prior art, the present invention proposes a method and system for monitoring the production quality of home textile fabrics based on industrial data. The method and system aim to achieve early warning of the quality of home textile fabrics by real-time monitoring of various data in the production process, and optimize equipment parameters, thereby improving production quality.

[0023] The specific application of the present invention will be described in detail below through two embodiments.

[0024] Embodiment 1: In the embodiment of the present application, the production quality of home textile fabric A is monitored; wherein the home textile fabric A is a pure cotton bedding fabric; the home textile fabric A is monitored by the method and system of the present invention; Figure 1 As shown, in Figure 1The specific implementation process of the method of the present invention is given in the following, including: S10. Obtaining the planned finished product information of the home textile fabric to be processed; S20. Obtaining the initial equipment parameters according to the planned finished product information; S30. Setting the equipment using the initial equipment parameters; S40. Taking the home textile fabric to be processed as the target object for production; S50. Using the home textile fabric quality monitoring model to monitor the quality of the target object; S60. Judging whether there is a quality problem in the monitoring result; Figure 2 The system structure of the present invention is given in the figure, including: a home textile fabric production plan generation unit, a template unit, an initial equipment parameter generation unit, a production unit, a home textile fabric quality monitoring unit, an equipment parameter optimization unit and a feedback unit; this application example will be combined with Figure 1 and Figure 2 Provide detailed explanation; According to the home textile fabric production plan generation unit of the system, the planned finished product information of home textile fabric A is obtained, corresponding to Figure 1 Step S10; wherein the planned finished product information includes: production specifications, models, quantities, batches, and process requirements, etc.; Further, the initial equipment parameter generation unit of the system is used to obtain the initial equipment parameters of each equipment in the production process of home textile fabric A according to the planned finished product information, corresponding to step S20; wherein, refer to step S21. Figure 3 ,The specific implementation process of the initial device parameter generation unit includes: According to the planned finished product information, the corresponding home textile fabric template is matched from the template library managed by the template unit; wherein, the selection of the home textile fabric template is mainly through keyword matching between the planned finished product information and the template finished product information, and the matching process at least includes keywords such as model and process requirements; Further, the difference between the planned finished product information and the home textile fabric template is calculated; wherein the difference refers to the difference in size and production process between the production specifications in the planned finished product information of home textile fabric A and the matched home textile fabric template; by comparing the difference in size between home textile fabric A and the home textile fabric template; For example, the calculation formula for size difference is: ;in, To plan the finished product size; is the template size; the calculation formula for the production process difference is: ;in, is the total number of production process items; For the planned finished product Production process vectors; For the template Production process vectors; When the difference is 0, the standard equipment parameters corresponding to the selected home textile fabric template are used as the initial equipment parameters for the production of home textile fabric A; When the difference is not 0, the deviation value between the home textile surface A and the home textile fabric template is calculated, and a home textile fabric deviation matrix is ​​constructed; wherein the deviation value refers to the deviation value of each item between the planned finished product information and the template finished product information, including: size deviation value, production process deviation value and fabric property deviation value, etc.; the deviation values ​​listed above are only partial descriptions of the embodiments of this application, and the deviation calculation can be performed specifically according to the content of the finished product information.

[0025] Further, according to the deviation value in the home textile fabric deviation matrix, the equipment attribute of each production step is obtained; wherein the equipment attribute refers to the equipment name; Furthermore, a device-parameter mapping relationship between device attributes and standard device parameters in the home textile fabric template is established. For details about the device-parameter mapping, please refer to the description of some mapping columns given in Table 1.

[0026] Table 1 Device--parameter mapping part example

[0027] Furthermore, the home textile fabric deviation matrix and the equipment-parameter mapping are input into the equipment parameter correction function; wherein the equipment parameter correction function is: ;in, Corrected parameters for standard equipment; are standard equipment parameters; The mapping matrix corresponding to the device-parameter mapping; is the deviation matrix for home textile fabrics; is a nonlinear correction function; For production environment data; Among them, the nonlinear correction function is expressed as: ;in, is the nonlinear relationship between equipment parameters and fabric deviation; is the nonlinear relationship between equipment parameters and fabric deviation; Linear correction function ;in, To standardize operations; and is a constant used to adjust the strength of the nonlinear relationship; ;in, , and is a constant used to adjust the impact of production environment data on the correction process; in, , , , , and It can be obtained by fitting historical data; Furthermore, the standard equipment correction parameters are used as initial equipment parameters.

[0028] Taking the content in Table 1 as an example, after correction by the equipment parameter correction function, the weaving density deviation of the loom is +2 / cm. According to the equipment-parameter mapping relationship, the standard weaving density of the loom is 80 / cm. After the deviation of +2 / cm, the adjusted weaving density should be 82 / cm. The dye concentration deviation of the dyeing machine is +1%. The original standard dye concentration is 5%, which should be adjusted to 6% after correction.

[0029] The embodiment of the present application adopts a method to set initial parameters for each production equipment of home textile fabric A; the method selects a home textile fabric template from a template library and adjusts the equipment parameters according to the difference between the planned finished product information and the template. When the difference is 0, the standard equipment parameters of the template are directly used; when the difference is not 0, the deviation value is calculated and the deviation matrix is ​​constructed, and the adjustment is made through the equipment parameter correction function, and finally the initial equipment parameters suitable for home textile fabric A are obtained. The equipment parameters can be dynamically adjusted according to specific production needs, the accuracy and efficiency of the production process can be improved, and the consistency of the final product quality can be ensured.

[0030] Further, according to the content of step S30, the initial equipment parameters of each equipment on the production line are set; Further, according to the content of step S40, home textile fabric A is taken as a target object and produced by the production unit of the system; Furthermore, during the production process of the production unit, the home textile fabric quality monitoring unit of the system is used for monitoring, corresponding to the content of S50; wherein the home textile fabric quality monitoring unit uses the home textile fabric quality monitoring model for quality monitoring, and the specific process includes: using the target object image currently being produced after the acquisition layer; wherein the acquisition layer is to use high-definition monitoring equipment to collect images of the home textile fabric A in the production process in real time; Furthermore, the target object image is input into a preprocessing layer to obtain a standard target object image; wherein the preprocessing layer includes: performing background segmentation on the target object image to obtain a first target object image; performing denoising on the first target object image to obtain a second target object image; performing smoothing on the second target object image to obtain a third target object image; performing edge detection on the third target object image to obtain a fourth target object image; and performing image enhancement on the fourth target object image to obtain a standard target object image.

[0031] In the embodiment of the present application, a preprocessing layer is used to process the collected images, including: background segmentation, denoising, smoothing, edge detection and image enhancement; the above methods can remove noise, enhance image details, extract key features, and improve the accuracy of the monitoring model, thereby more accurately identifying unqualified products.

[0032] Further, the standard target object image is input into the recognition separation layer to obtain the bottom layer image and the top layer image; see Figure 4 , wherein the identification separation layer includes: starting from the upper left corner of the standard target object image, the identification frame is moved horizontally to obtain the identification frame image and the identification frame coordinates; wherein the identification frame adopts the sliding window technology; in the embodiment of the present application, the size of the identification frame is set to 64×64 pixels, the interval of each movement is set to 32 pixels, and the movement is performed every 2 frames; Extract features of the recognition frame image to obtain a first feature vector; The first feature vector is input into the classification model to obtain the classification result; wherein the classification model is a trained convolutional neural network (CNN) for performing a binary classification operation on the image in the recognition frame, and the classification result includes two types: bottom layer and top layer; The standard target object image is cropped according to the classification result to obtain the bottom layer image and the top layer image.

[0033] In the embodiment of the present application, the results of multiple rounds of classification show that the parameters set for the above-mentioned recognition frame are optimal values; wherein, in this process, the present invention also sets multiple recognition frame parameters, such as: (1) recognition frame size: 64×64 pixels, each movement interval: 32 pixels, update frequency: update every frame; (2) recognition frame size: 64×64 pixels, each movement interval: 64 pixels, update frequency: update every 2 frames; (3) recognition frame size: 64×64 pixels, each movement interval: 128 pixels, update frequency: update every 3 frames; (4) recognition frame size: 128×128 pixels, each movement interval: 32 pixels, update frequency: update every frame.

[0034] Refer to Table 2. According to the parameters set by the recognition box, the classification accuracy comparison obtained by the classification model is as follows: Table 2 Comparison of classification accuracy of different recognition box parameters

[0035] Recognition box size: 64×64 pixels, moving interval 64 pixels, update every 2 frames (accuracy: 88.5%) This is the setting with the highest classification accuracy in the table. Compared with the first setting, the moving interval is increased to 64 pixels and updated every 2 frames, which means that the step size of each slide is increased, but the update frequency is reduced. The advantage of this is that the computational burden is reduced, and the key features of the image can still be captured well. The results show that this setting has a classification accuracy of 88.5%, which is considered to be the best parameter combination.

[0036] The last setting uses a larger recognition box (128×128 pixels) and uses per-frame updates and a smaller movement interval (32 pixels). Although increasing the size of the recognition box helps capture more global information, the corresponding amount of calculation is also greatly increased. Larger recognition boxes lead to higher calculation and memory consumption, making the model processing slower, and the high frequency of per-frame updates did not improve the classification accuracy, which eventually caused the accuracy to drop to 80.7%.

[0037] Further, the bottom layer image and the top layer image are input into the quality monitoring layer to obtain the quality monitoring result; wherein the quality monitoring layer includes: a bottom layer quality monitoring sublayer, a top layer quality monitoring sublayer and a joint quality monitoring sublayer; wherein the specific process of the quality monitoring layer includes: inputting the bottom layer image into the bottom layer quality monitoring sublayer to obtain the bottom layer features of the target object; wherein the bottom layer features of the target object include: bottom layer specification features, bottom layer structure features, bottom layer color features and bottom layer surface features; Perform bottom-level quality analysis on the bottom-level features of the target object to obtain the bottom-level fabric qualification vector; the bottom-level fabric qualification vector includes: bottom-level specification qualification, bottom-level structure qualification, bottom-level color qualification and bottom-level defect qualification; when any qualification in the bottom-level fabric qualification vector does not meet the bottom-level standard, mark the bottom-level unqualified type and obtain the corresponding unqualified bottom-level data; Input the top-level image into the top-level quality monitoring sublayer to obtain the top-level features of the target object; wherein the top-level features of the target object include: top-level shape features, top-level color features, top-level texture features, and top-level integrity features; Perform top-level quality analysis on the top-level features of the target object to obtain the top-level quality vector of the fabric; wherein the top-level quality vector of the fabric includes: top-level shape quality, top-level color quality, top-level texture quality and top-level integrity quality; when any quality item in the top-level quality vector of the fabric does not meet the top-level standard, mark the top-level unqualified type and obtain the corresponding unqualified top-level data; Input the bottom layer image and the top layer image into the joint quality monitoring sublayer, and perform image superposition according to the standard target object image to obtain the target object joint image; Extracting joint features of the joint image of the target object; wherein the joint features include: position features, size features and scale features; Calculate the expected error between the joint feature and the current production expectation, and construct the production error vector; when the comprehensive error value of the production error vector exceeds the error tolerance, mark the joint unqualified type and obtain the corresponding unqualified joint data; by analyzing the fluctuation range and distribution of errors in historical data, a common error range can be obtained. This range can be used to set the error tolerance.

[0038] In the embodiment of the present application, the quality monitoring layer is used to perform multi-level quality monitoring on the bottom layer, top layer and joint image of the home textile fabric A, so as to comprehensively evaluate the specifications, structure, color, texture and other characteristics of the fabric, timely discover unqualified items and obtain detailed unqualified parameters, which helps to ensure that the product meets the production standards and improve the accuracy of the production process and product quality.

[0039] Furthermore, the output layer is used to output the quality monitoring results; wherein, according to the monitoring results of the quality monitoring layer, "the bottom structure is unqualified" and "the top color is unqualified" are obtained.

[0040] Further, the equipment parameter optimization unit is used to optimize the initial equipment parameters of the current production step, corresponding to step S60; during the optimization process, it is necessary to determine whether optimization is required based on the monitoring results; when the monitoring results do not have quality problems, the initial equipment parameters of the current production step; when the monitoring results have quality problems, the initial equipment parameters of the current production step are adjusted as equipment adjustment parameters, and quality monitoring is continuously performed; According to the results of the home textile fabric quality monitoring model, the initial equipment parameters of the current production step are adjusted; the specific process is: obtaining the unqualified type of the target object and the corresponding unqualified data; refer to Table 3, and the corresponding data description is given in Table 3; Table 3 Description of quality monitoring results

[0041] Further, the unqualified type and the unqualified data are characterized to obtain an unqualified feature vector; Further, the unqualified feature vector and the initial equipment parameters are input into a random forest model to obtain an equipment parameter influence vector; Further, a correlation analysis is performed on the equipment parameter influence vector to obtain an equipment parameter correlation matrix; wherein the equipment parameter correlation matrix is ​​calculated using a Pearson correlation coefficient analysis method; Further, the current production expectation is obtained, and the current production expectation feature vector is obtained by characterization; Furthermore, the unqualified feature vector, the initial equipment parameters and the equipment parameter association matrix are input into the deduction model; wherein the deduction model is a machine learning algorithm that automatically identifies the complex relationship between equipment parameters and product quality based on a large amount of historical production data and equipment operation data. The model can analyze the gap between the unqualified feature vector and production expectations in real time, and adjust the initial equipment settings according to the equipment parameter association matrix. By continuously optimizing these input parameters, the deduction model can predict and optimize equipment parameters; The function of the deduction model is defined as: ;in, is the deduction model function; For the current production step Initial device parameters; is the unqualified feature vector; is the device parameter association matrix; For the current production step The unqualified feature vector is obtained by deduction; The expected feature vector for the current production.

[0042] When the error between the deduced unqualified feature vector and the current production expected feature vector is the smallest, the initial equipment parameters are updated and the equipment adjustment parameters of the current production step are obtained.

[0043] The embodiment of the present application adjusts the equipment parameters of the current production step according to the quality monitoring results, extracts the unqualified feature vector by analyzing the unqualified data, and inputs it into the random forest model together with the initial equipment parameters to obtain the equipment parameter influence vector. Next, the equipment parameter correlation analysis is performed to construct the equipment parameter correlation matrix. By characterizing and comparing with the current production expectations, the model is deduced to optimize the equipment parameters until the error between the unqualified feature vector and the expected feature vector is minimized, thereby obtaining the updated equipment adjustment parameters, which can greatly improve the quality, stability and efficiency of production.

[0044] Furthermore, the feedback unit of the system is used to provide feedback on monitoring results and equipment parameter adjustments during the production process.

[0045] In the embodiment of the present application, the equipment parameters in the production process of home textile fabric A are dynamically adjusted and the quality is monitored; specifically, first, by selecting a suitable home textile fabric template from the template library and correcting the production parameters according to the difference and deviation matrix, the production requirements and equipment configuration can be accurately matched, and production errors caused by improper equipment parameter settings can be avoided, ensuring that each link in the production process is running under the optimal parameters. Secondly, the quality monitoring model uses image processing and machine learning algorithms to carefully analyze the bottom and top features of home textile fabrics, so as to accurately identify quality problems that may occur in the production process. By performing quality analysis on the multi-dimensional features of the bottom and top layers (such as color, structure, shape, etc.), defects can be discovered in time at the beginning of production, and corrected by adjusting the equipment parameters, avoiding the production of a large number of defective products and improving the consistency and reliability of the product. In addition, the adaptive ability of the random forest model and the deduction model in the embodiment of the present application in the case of quality abnormalities can flexibly adjust the equipment parameters according to the deviation information in the actual production process to adapt to different production requirements. When there are quality problems in the monitoring results, the model automatically optimizes the equipment parameters by analyzing the correlation between the unqualified feature vectors and the equipment parameters, and feeds the optimized parameters back to the production system, thereby achieving continuous quality monitoring and adjustment and forming a closed-loop control.

[0046] Embodiment 2: In Example 1, the quality monitoring of home textile fabric A during the production process and the control of production equipment parameters are realized; in order to illustrate that the present invention has good adaptability to production monitoring of different types of home textile fabrics; the present application example will explain the process of home textile fabric B as follows: Get the planned finished product information of home textile fabric B; Further, according to the planned finished product information, initial equipment parameters are obtained; wherein the difference degree of the home textile fabric template selected from the template library according to the finished product information is 0; therefore, the standard equipment parameters of the home textile fabric template are used as the initial equipment parameters of home textile fabric B; Furthermore, a home textile fabric quality monitoring model is used to perform quality monitoring on the target object; according to the quality monitoring result, it can be known that the home textile fabric B has a quality problem of "inconsistent thickness" in the weaving step, and unqualified thickness data of the home textile fabric B in this step is obtained; Further, the unqualified type of home textile fabric B is determined as "inconsistent thickness" and the corresponding unqualified data; Further, the unqualified type and unqualified data are characterized to obtain an unqualified feature vector; Furthermore, the unqualified feature vector and the initial equipment parameters are input into the random forest model to obtain the equipment parameter influence vector; Furthermore, the correlation analysis of the equipment parameter influence vectors is performed to obtain the equipment parameter correlation matrix; Further, the current production expectation is obtained, and the current production expectation feature vector is obtained by characterization; Furthermore, the unqualified feature vector, the initial equipment parameters and the equipment parameter association matrix are input into the deduction model; when the error between the unqualified feature vector and the current production expected feature vector is minimized, the initial equipment parameters are updated and the equipment adjustment parameters of the current production step are obtained.

[0047] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for monitoring the production quality of home textile fabrics based on industrial data, characterized in that: include: Obtain the planned finished product information of the home textile fabrics to be processed; According to the planned finished product information, obtaining initial equipment parameters; Setting the device using the initial device parameters; The home textile fabric to be processed is produced as a target object; The home textile fabric quality monitoring model is used to monitor the quality of the target object, including: an acquisition layer for acquiring the target object image during the monitoring process; a preprocessing layer for performing image processing on the target object image to obtain a standard target object image; The identification and separation layer is used to identify and separate the bottom layer and the top layer of the standard target object image. The specific process includes: starting from the upper left corner of the standard target object image through the identification frame, moving horizontally to obtain the identification frame image and the identification frame coordinates; extracting the features of the identification frame image to obtain a first feature vector; inputting the first feature vector into the classification model to obtain a classification result; cropping the standard target object image according to the classification result to obtain a bottom layer image and a top layer image; Quality monitoring layer: used to monitor the quality of the bottom layer image and the top layer image; An output layer, used to output the monitoring results of the quality monitoring layer; When there are no quality problems with the monitoring results, the initial equipment parameters of the current production step are used; when there are quality problems with the monitoring results, the initial equipment parameters of the current production step are adjusted as equipment adjustment parameters, and quality monitoring is continuously performed.

2. The method for monitoring the production quality of home textile fabrics based on industrial data according to claim 1, characterized in that: The process of obtaining the initial equipment parameters includes: selecting a home textile fabric template from a template library according to the planned finished product information; calculating the difference between the planned finished product information and the home textile fabric template; When the difference is 0, the standard equipment parameters of the home textile fabric template are used as the initial equipment parameters of the home textile fabric to be processed; When the difference is not 0, the deviation value between the home textile fabric to be processed and the home textile fabric template is calculated, and a home textile fabric deviation matrix is ​​constructed; according to the deviation value in the home textile fabric deviation matrix, the equipment attributes of each production step are obtained; Establishing a device-parameter mapping between the device attributes of the home textile fabric template and the standard device parameters; obtaining standard device correction parameters through a device parameter correction function; wherein the device parameter correction function is: ;in, Corrected parameters for standard equipment; is the standard equipment parameter; A mapping matrix corresponding to the device-parameter mapping; is the home textile fabric deviation matrix; is a nonlinear correction function; The standard equipment correction parameters are used as the initial equipment parameters.

3. The method for monitoring the production quality of home textile fabrics based on industrial data according to claim 1, characterized in that: The preprocessing layer includes: performing background segmentation on the target object image to obtain a first target object image; performing denoising on the first target object image to obtain a second target object image; performing smoothing on the second target object image to obtain a third target object image; performing edge detection on the third target object image to obtain a fourth target object image; and performing image enhancement on the fourth target object image to obtain the standard target object image.

4. The method for monitoring the production quality of home textile fabrics based on industrial data according to claim 1, characterized in that: The quality monitoring layer includes: a bottom quality monitoring sublayer, a top quality monitoring sublayer and a joint quality monitoring sublayer; wherein the specific process of the quality monitoring layer includes: inputting the bottom image into the bottom quality monitoring sublayer to obtain the bottom features of the target object; performing bottom quality analysis on the bottom features of the target object to obtain the bottom fabric qualification vector; when any qualification in the bottom fabric qualification vector does not meet the bottom standard, marking the bottom unqualified type and obtaining the corresponding unqualified bottom data; The top layer image is input into the top layer quality monitoring sublayer to obtain the top layer features of the target object; the top layer features of the target object are subjected to top layer quality analysis to obtain the top layer qualification vector of the fabric; when any qualification item in the top layer qualification vector of the fabric does not meet the top layer standard, the top layer unqualified type is marked, and the corresponding unqualified top layer data is obtained; The bottom layer image and the top layer image are input into the joint quality monitoring sublayer, and the images are superimposed according to the standard target object image to obtain a joint image of the target object; the joint features of the joint image of the target object are extracted; the expected error between the joint features and the current production expectation is calculated to construct a production error vector; when the comprehensive error value of the production error vector exceeds the error tolerance, the joint unqualified type is marked, and the corresponding unqualified joint data is obtained.

5. The method for monitoring the production quality of home textile fabrics based on industrial data according to claim 1, characterized in that: When there are quality problems in the monitoring results, the process of adjusting the initial equipment parameters of the current production step includes: Obtain the unqualified type and corresponding unqualified data of the target object; characterize the unqualified type and the unqualified data to obtain an unqualified feature vector; input the unqualified feature vector and the initial equipment parameters into a random forest model to obtain an equipment parameter influence vector; perform correlation analysis on the equipment parameter influence vector to obtain an equipment parameter association matrix; obtain the current production expectation and obtain the current production expectation feature vector through characterization; input the unqualified feature vector, the initial equipment parameters and the equipment parameter association matrix into a deduction model; when the error between the deduced unqualified feature vector and the current production expectation feature vector is minimized, the initial equipment parameters are updated to obtain the equipment adjustment parameters of the current production step.

6. A home textile fabric production quality monitoring system based on industrial data, characterized in that: The system comprises: a home textile fabric production plan generating unit, which is used to generate a generation plan for home textile fabrics to be processed; a template unit, which is used to store templates of various types of home textile fabrics, as well as corresponding template information and standard equipment parameters; an initial equipment parameter generating unit, which is used to generate initial parameters of various equipment in the production process of home textile fabrics to be processed; a production unit, which is used to produce home textile fabrics to be processed; a home textile fabric quality monitoring unit, which is used to monitor the quality of home textile fabrics to be processed during the production process; an equipment parameter optimization unit, which is used to optimize the initial parameters of the equipment; and a feedback unit, which is used to provide real-time feedback on the quality monitoring results of the home textile fabrics to be processed and the equipment parameter adjustment status.

7. The home textile fabric production quality monitoring system based on industrial data according to claim 6 is characterized in that: The initial equipment parameter generation unit comprises: selecting a home textile fabric template from a template library according to the planned finished product information; calculating the difference between the planned finished product information and the home textile fabric template; When the difference is 0, the standard equipment parameters of the home textile fabric template are used as the initial equipment parameters of the home textile fabric to be processed; When the difference is not 0, the deviation value between the home textile fabric to be processed and the home textile fabric template is calculated, and a home textile fabric deviation matrix is ​​constructed; according to the deviation value in the home textile fabric deviation matrix, the equipment attributes of each production step are obtained; Establishing a device-parameter mapping between the device attributes of the home textile fabric template and the standard device parameters; obtaining standard device correction parameters through a device parameter correction function; wherein the device parameter correction function is: ;in, Corrected parameters for standard equipment; is the standard equipment parameter; A mapping matrix corresponding to the device-parameter mapping; is the home textile fabric deviation matrix; is a nonlinear correction function; The standard equipment correction parameters are used as the initial equipment parameters.

8. The home textile fabric production quality monitoring system based on industrial data according to claim 6 is characterized in that: The home textile fabric quality monitoring unit adopts a home textile fabric quality monitoring model to perform quality monitoring, including: taking the home textile fabric to be processed in the production unit as the target object; using the acquisition layer to obtain the target object image; inputting the target object image into the preprocessing layer to obtain the standard target object image; inputting the standard target object image into the recognition separation layer to obtain the bottom layer image and the top layer image; wherein the recognition separation layer includes: starting from the upper left corner of the standard target object image through the recognition frame, moving horizontally to obtain the recognition frame image and the recognition frame coordinates; extracting the features of the recognition frame image to obtain the first feature vector; inputting the first feature vector into the classification model to obtain the classification result; cropping the standard target object image according to the classification result to obtain the bottom layer image and the top layer image; inputting the bottom layer image and the top layer image into the quality monitoring layer to obtain the quality monitoring result; and outputting the quality monitoring result using the output layer.

9. The home textile fabric production quality monitoring system based on industrial data according to claim 8 is characterized in that: The preprocessing layer includes: performing background segmentation on the target object image to obtain a first target object image; performing denoising on the first target object image to obtain a second target object image; performing smoothing on the second target object image to obtain a third target object image; performing edge detection on the third target object image to obtain a fourth target object image; and performing image enhancement on the fourth target object image to obtain the standard target object image.

10. The home textile fabric production quality monitoring system based on industrial data according to claim 8, characterized in that: The specific process of the quality monitoring layer includes: Input the bottom layer image into the bottom layer quality monitoring sublayer to obtain the bottom layer features of the target object; perform bottom layer quality analysis on the bottom layer features of the target object to obtain the bottom layer qualification vector of the fabric; when any qualification in the bottom layer qualification vector of the fabric does not meet the bottom layer standard, mark the bottom layer unqualified type and obtain the corresponding unqualified bottom layer data; The top layer image is input into the top layer quality monitoring sublayer to obtain the top layer features of the target object; the top layer features of the target object are subjected to top layer quality analysis to obtain the top layer qualification vector of the fabric; when any qualification item in the top layer qualification vector of the fabric does not meet the top layer standard, the top layer unqualified type is marked, and the corresponding unqualified top layer data is obtained; The bottom layer image and the top layer image are input into the joint quality monitoring sublayer, and the images are superimposed according to the standard target object image to obtain a joint image of the target object; the joint features of the joint image of the target object are extracted; the expected error between the joint features and the current production expectation is calculated to construct a production error vector; when the comprehensive error value of the production error vector exceeds the error tolerance, the joint unqualified type is marked, and the corresponding unqualified joint data is obtained.

Images