A method and system for real-time quality rating of yarn textile

Abstract

This application relates to the field of yarn textile manufacturing, and in particular to a real-time quality rating method and system for yarn textile manufacturing. The method includes: acquiring a spatiotemporal charge dataset; generating a yarn electrostatic state dataset through a multimodal sensing fusion strategy; identifying yarn textile quality defects based on the yarn electrostatic state dataset and real-time acquired yarn morphology data, and performing a two-dimensional dynamic rating based on the defect type and severity to generate a real-time yarn quality rating; analyzing the changing trend of the real-time yarn quality rating to generate a subsequent quality change trend; executing corresponding process control instructions based on the subsequent quality change trend, and outputting a quality rating log. This application enables proactive process optimization and improves quality traceability in the yarn textile manufacturing process, enhancing the accuracy and real-time performance of quality rating.

Description

Technical Field

[0001] This application relates to the field of yarn textiles, and in particular to a real-time quality rating method and system for yarn textiles. Background Technology

[0002] In the textile industry, yarn, as the basic unit of fabric production, constitutes the core lifeline of the trillion-dollar apparel, home textiles and industrial textiles manufacturing system, and is the foundation for supporting the efficient operation and value realization of the huge textile industry chain.

[0003] However, existing yarn quality testing technologies cannot accurately detect and quantify the dynamic details of high-speed running yarns during the production process. This leads to difficulty in timely intervention of quality fluctuations, continuous production of defective products, and increased waste of raw materials, ultimately weakening product competitiveness and increasing production costs. Summary of the Invention

[0004] This application provides a real-time quality rating method and system for yarn textiles to solve the above-mentioned technical problems.

[0005] In a first aspect, this application provides a real-time quality rating method for yarn textiles, the method comprising:

[0006] A spatiotemporal charge dataset is acquired, and a yarn electrostatic state dataset is generated using a multimodal sensing fusion strategy. Based on this dataset and real-time acquired yarn morphology data, yarn textile quality defects are identified, and a two-dimensional dynamic rating is performed according to the defect type and severity to generate a real-time yarn quality rating. The real-time quality rating is then analyzed for its changing trends to generate a subsequent quality change trend. Based on this trend, corresponding process control instructions are executed, and a quality rating log is output.

[0007] This solution first collects spatiotemporal charge data, then generates a yarn electrostatic state dataset through multimodal sensor fusion, achieving precise quantification of electrostatic factors. It then integrates electrostatic state and morphological data to extract electrostatic characteristics associated with defects, generating real-time quality results through dual-dimensional dynamic rating, improving the comprehensiveness and accuracy of defect identification. By combining yarn category information and electrostatic correlation weights, it calibrates electrostatic data and predicts quality change trends, enhancing the targeted nature of the analysis. Based on these trends, it executes forward-looking process adjustments and outputs quality rating logs to users, forming a closed-loop management system. By generating an electrostatic state dataset through multimodal sensor fusion, identifying defects and assigning dual-dimensional ratings based on morphological data, and then predicting trends and adjusting accordingly based on category information and electrostatic weights, it can accurately detect and quantify the dynamic details of high-speed running yarn in real time. This allows for timely intervention in quality fluctuations, reduces the continuous production of defective products, avoids further waste of raw materials, ultimately enhances product competitiveness and reduces production costs.

[0008] Optionally, the multimodal sensing fusion strategy includes: using a spatially deployed ring array of electrostatic sensors to collect spatiotemporal charge datasets on key points of the yarn in real time without contact; the spatiotemporal charge dataset includes real-time charge intensity values ​​and real-time charge distribution characteristics; using a high frame rate visual sensor to capture yarn morphology data and motion trajectory in real time; simultaneously analyzing the spatiotemporal correlation between the real-time charge intensity values, the real-time charge distribution characteristics, and the yarn morphology data to identify abnormal electrostatic adsorption trajectories caused by quality defects; and generating the yarn electrostatic state dataset based on the spatiotemporal charge dataset, the yarn morphology data, and the abnormal electrostatic adsorption trajectories.

[0009] Optionally, the step of identifying yarn textile quality defects based on the yarn electrostatic state dataset and combined with real-time collected yarn morphology data, and performing a two-dimensional dynamic rating based on the defect type and severity to generate an initial yarn quality rating, includes: identifying yarn defect types based on the yarn electrostatic state dataset and the real-time collected yarn morphology data, and determining the severity of the corresponding explicit defects for each yarn defect type; the yarn defect types include excessive hairiness, uneven yarn count, and fiber knots; normalizing several yarn defect types within a unified numerical range to determine the normalized yarn defect types; determining a dynamic rating score based on the normalized yarn defect types and the severity of the explicit defects; and generating a real-time yarn quality rating based on the dynamic rating score, wherein the real-time yarn quality rating is divided into three levels: excellent, medium, and poor.

[0010] Optionally, the step of identifying yarn defect types and determining the severity of visible defects corresponding to the yarn defect types based on the yarn electrostatic state dataset and the real-time acquired yarn morphology data includes: acquiring a real-time image of the yarn appearance and the corresponding real-time charge intensity value; and performing the following hair density gridding analysis steps based on the yarn appearance image: when the number of discrete fibers in several consecutive detection data exceeds a preset first density threshold and the average electrostatic intensity is greater than a preset multiple of the visual hair count, it is determined to be a local hair aggregation defect; when the number of discrete fibers in a unit detection area exceeding a preset proportion exceeds a preset second density threshold and the overall electrostatic distribution non-uniformity is greater than a preset diffuse hair proportion, it is determined to be a diffuse hair excess defect; and analyzing the spatial distribution state of hair and quantifying the degree of hair density visibility based on the combination pattern of the number of local aggregation areas and the diffuse distribution proportion, combined with the coupling coefficient between the electrostatic intensity increase and hair density.

[0011] Optionally, the step of identifying yarn defect types and determining the severity of explicit defects corresponding to the yarn defect types based on the yarn electrostatic state dataset and the real-time acquired yarn morphology data includes: acquiring a yarn axial image sequence in real time, and performing the following three-dimensional fluctuation analysis steps based on the yarn axial image sequence: marking abnormal diameter regions where the diameter difference between adjacent sampling points exceeds the diameter tolerance threshold as abrupt unevenness regions; simultaneously detecting the rate of change of the electrostatic adsorption trajectory direction angle of the abnormal diameter regions; and marking the current abnormal diameter region as an electrostatic turbulence region when the rate of change of the direction angle exceeds a preset turbulence threshold. The flow pattern is abruptly uneven; the abnormal trend region corresponding to the unidirectional continuous increase / decrease in diameter exceeding a preset number of points is marked as a gradual uneven region. The slope of the electrostatic gradient distribution of the gradual region is obtained simultaneously. When the slope is in the same direction as the diameter change trend and exceeds the associated threshold, it is marked as an electrostatic accumulation gradual uneven region. The proportion of the length of the abruptly uneven region and the length of the gradually uneven region in the unit detection area is determined. The degree of unevenness is quantified by combining the maximum diameter deviation value positively correlated with the degree of unevenness and the electrostatic disturbance intensity parameter, and the electrostatic gradient consistency parameter negatively correlated with the degree of unevenness.

[0012] Optionally, the step of identifying yarn defect types and determining the severity of visible defects corresponding to the yarn defect types based on the yarn electrostatic state dataset and the real-time acquired yarn morphology data includes: acquiring real-time scanning imaging data of the yarn surface and simultaneously capturing the corresponding electrostatic intensity extreme values, and performing the following impurity multi-scale feature extraction steps to clarify impurity attributes: when there is a dark clump-like structure with irregular boundaries and a charge decay rate less than a preset decay rate threshold, the dark clump-like structure is determined to be a fiber agglomerate; when there is a reflective particle with geometric edges and an electrostatic intensity extreme value greater than a preset multiple of the particle size, the reflective particle is determined to be a hard impurity; when there is a radial fiber entanglement structure with an oscillating electrostatic intensity, the radial fiber entanglement structure is determined to be a yarn defect nodule; according to the size level, optical contrast, and boundary sharpness of the impurity multi-scale features, the size level, optical contrast, and boundary sharpness are weighted according to the impurity attributes to quantify the degree of fiber agglomeration visibility.

[0013] Optionally, the step of generating a real-time yarn quality rating based on the dynamic rating score, wherein the real-time yarn quality rating is divided into three levels: excellent, medium, and poor, includes: mapping several identified yarn defect types to defect type identifiers based on a preset rule table, and mapping the assessed severity of the explicit defects to severity level identifiers; concatenating several defect type identifiers with their corresponding severity level identifiers, and using the concatenated result as a unit rating code; after all the unit rating codes are concatenated, performing a second string concatenation on several unit rating codes to generate a composite rating code; and mapping the composite rating code to an initial quality level of excellent / medium / poor according to a preset mapping rule; the mapping... The rules include: determining the total number of all defects as the defect density, and identifying the highest severity level identifier in the composite rating code; if the defect density is greater than a first density threshold or the highest severity level identifier reaches the first severity level threshold, it is judged as poor; if the defect density does not meet the aforementioned condition for judging as poor but is greater than a second density threshold or the highest severity level identifier reaches the second severity level threshold, it is judged as medium; if the defect density does not exceed the second density threshold and the highest severity level identifier does not reach the second severity level threshold, it is judged as excellent; wherein, the first density threshold is greater than the second density threshold, and the first severity level threshold is higher than the second severity level threshold.

[0014] Optionally, the step of analyzing the trend of the real-time quality rating of the yarn and generating a subsequent quality change trend of the yarn based on the real-time quality rating of the yarn includes: extracting the trajectory temporal dynamic features, including trajectory fluctuation frequency, directional consistency, and intensity variation amplitude, based on the electrostatic adsorption abnormal trajectory; identifying trajectory evolution patterns based on the dynamic features: when the trajectory fluctuation frequency is lower than a preset low-frequency threshold and the directional consistency is higher than a stable threshold, it is a stable and gradual mode; when the trajectory fluctuation frequency is higher than a preset high-frequency threshold and the intensity variation amplitude continues to increase, it is a deterioration and diffusion mode; when the directional consistency is lower than a disorder threshold and the intensity variation amplitude decreases periodically, it is an intermittent oscillation mode; mapping the trajectory evolution patterns to quality trends: the stable and gradual mode is mapped to a defect mitigation trend, the deterioration and diffusion mode is mapped to a defect escalation trend, and the intermittent oscillation mode is mapped to a quality fluctuation trend; and generating a subsequent quality change trend of the yarn based on the quality trends and the severity of the explicit defects.

[0015] Optionally, the step of executing corresponding process control instructions and outputting a quality rating log based on the subsequent quality change trend of the yarn includes: performing differentiated control based on the type of the subsequent quality change trend of the yarn: if it is determined to be a defect escalation trend, triggering linkage control to reduce spinning speed and increase environmental humidification; if it is determined to be a defect mitigation trend, maintaining the current process and reducing detection density; if it is determined to be a quality fluctuation trend, dynamically adjusting environmental temperature and humidity and directionally initiating static electricity neutralization; associating the real-time quality rating of the yarn, the subsequent quality change trend of the yarn, and the control instructions with production line batches and time, generating and outputting a quality rating log.

[0016] Secondly, this application provides a real-time quality rating system for yarn textiles, the system comprising:

[0017] The charge monitoring module is used to acquire spatiotemporal charge datasets and generate yarn electrostatic state datasets through a multimodal sensing fusion strategy.

[0018] The dual-dimensional rating module is used to identify yarn textile quality defects based on the yarn electrostatic state dataset and real-time collected yarn morphology data, and to perform dual-dimensional dynamic rating according to the defect type and severity to generate a real-time yarn quality rating.

[0019] The trend projection module is used to analyze the changing trend of the real-time quality rating of the yarn based on the real-time quality rating of the yarn, and generate the subsequent quality change trend of the yarn.

[0020] The control execution module is used to execute corresponding process control instructions based on the subsequent quality change trend of the yarn, and output a quality rating log. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a schematic diagram illustrating an application scenario provided in one embodiment of this application;

[0023] Figure 2 A flowchart illustrating a real-time quality rating method for yarn spinning, provided as an embodiment of this application;

[0024] Figure 3 This is a schematic diagram of a real-time quality rating system for yarn spinning provided in an embodiment of this application. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0026] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article, unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship.

[0027] The embodiments of this application will now be described in further detail with reference to the accompanying drawings.

[0028] However, existing yarn quality testing technologies cannot accurately detect and quantify the dynamic details of high-speed running yarns during the production process. This leads to difficulty in timely intervention of quality fluctuations, continuous production of defective products, and increased waste of raw materials, ultimately weakening product competitiveness and increasing production costs.

[0029] Based on this, this application provides a real-time quality rating method and system for yarn textile manufacturing. First, spatiotemporal charge data is collected. A yarn electrostatic state dataset is generated through multimodal sensor fusion, enabling precise quantification of electrostatic factors. Electrostatic state data and morphological data are fused to extract electrostatic characteristics associated with defects. Real-time quality results are generated through dual-dimensional dynamic rating, improving the comprehensiveness and accuracy of defect identification. Combining yarn category information and electrostatic correlation weights, electrostatic data is calibrated and quality change trends are predicted, enhancing the targeted nature of the analysis. Based on these trends, forward-looking process control is implemented, and quality rating logs are output to the user, forming a closed-loop management system. By generating an electrostatic state dataset through multimodal sensor fusion, identifying defects and rating them in a dual-dimensional manner by combining morphological data, and then predicting trends and adjusting them based on category information and electrostatic weights, the system can accurately detect and quantify the dynamic details of high-speed running yarn in real time. This allows for timely intervention in quality fluctuations, reduces the continuous production of defective products, avoids further waste of raw materials, and ultimately enhances product competitiveness and reduces production costs.

[0030] Figure 1 This is a schematic diagram illustrating an application scenario provided by this application. In the yarn spinning process, this application enables forward-looking process optimization and improves quality traceability, thereby enhancing the accuracy and real-time nature of quality rating.

[0031] Specifically, the method of this application is applied to any server that communicates with an electrostatic induction ring array. The server acquires a spatiotemporal charge dataset provided by the array. First, spatiotemporal charge data is collected. Then, a yarn electrostatic state dataset is generated through multimodal sensing fusion, achieving precise quantification of electrostatic factors. Electrostatic state data and morphological data are fused to extract electrostatic characteristics associated with defects. Real-time quality results are generated through dual-dimensional dynamic rating, improving the comprehensiveness and accuracy of defect identification. Combining yarn category information and electrostatic correlation weights, electrostatic data is calibrated and quality change trends are predicted, enhancing the targeting of the analysis. Based on the trends, forward-looking process control is executed, and a quality rating log is output to the user, forming a closed-loop management system. Specific implementation methods can be found in the following embodiments.

[0032] Figure 2 This is a flowchart illustrating a real-time quality rating method for yarn spinning according to an embodiment of this application. The method of this embodiment can be applied to the server in the above scenario. Figure 2 As shown, the method includes:

[0033] S201. Obtain the spatiotemporal charge dataset and generate a yarn electrostatic state dataset through a multimodal sensing fusion strategy.

[0034] Spatiotemporal charge datasets refer to collections of electrostatic charge data collected at different spatial locations over time during yarn production. This includes real-time charge intensity values ​​and real-time charge distribution characteristics, and the data originates from a ring array of electrostatic sensors surrounding the yarn. Multimodal sensor fusion strategies can be a processing strategy that integrates different types of sensor data from the ring array of electrostatic sensors and high-frame-rate visual sensors.

[0035] A yarn electrostatic state dataset can be used to characterize the charge data, morphological data, and abnormal electrostatic adsorption trajectories of yarn during the textile process, and can intuitively reflect the electrostatic characteristics of yarn.

[0036] Specifically, in the yarn spinning process, static electricity is a key, yet often overlooked, factor affecting quality. Existing technologies rely heavily on manual visual inspection or single-form sensors, only capable of capturing macroscopic defects in the yarn's appearance (such as yarn breakage). They lack the ability to accurately monitor and quantitatively assess the microscopic dynamics of high-speed yarns in real time, creating blind spots in the production process. This leads to difficulty in timely correction of quality anomalies, increased defect rates, and increased raw material losses, ultimately hindering product market competitiveness and significantly increasing overall costs. Furthermore, data collected by single charge sensors is easily affected by environmental temperature and humidity, equipment vibration, etc., resulting in data bias and failing to accurately reflect the true state of yarn static electricity. This step deploys devices such as a ring array of static sensors at key locations on the spinning machine to collect real-time data on charge distribution and movement trajectories on the yarn surface and surrounding space, forming a spatiotemporal charge dataset. A multimodal sensing fusion strategy is then employed to extract static characteristic parameters, generating a yarn static state dataset. The acquisition of spatiotemporal charge datasets can comprehensively capture the dynamic changes of charge on the yarn surface and in the surrounding space. The multimodal sensing fusion strategy, by integrating multi-source data such as charge and optics, can eliminate the error of a single sensor and filter noise interference. The resulting yarn electrostatic state dataset can accurately quantify electrostatic characteristics, providing a reliable basis for subsequent quality analysis.

[0037] S202. Based on the yarn electrostatic state dataset and combined with the real-time collected yarn morphology data, identify yarn textile quality defects, and perform a two-dimensional dynamic rating according to the defect type and severity to generate a real-time yarn quality rating.

[0038] Yarn morphology data refers to a set of data reflecting the physical appearance and structural characteristics of yarn, including parameters such as yarn diameter, hair length and density, and evenness variation. Yarn textile quality defects refer to problems in the yarn that do not meet quality standards due to factors such as raw material characteristics, equipment parameters, and environmental conditions during the spinning process. Common types include excessive hairiness, unevenness, and fiber knots. Dual-dimensional dynamic rating is a method of real-time assessment of yarn quality from two dimensions: "defect type" and "severity." The defect type dimension categorizes different quality problems, while the severity dimension classifies defects according to their impact on subsequent processing and product performance (e.g., minor, moderate, severe). Real-time yarn quality rating is an assessment result of the current yarn quality level generated through dual-dimensional dynamic rating based on real-time collected electrostatic state and morphology data, directly reflecting the immediate quality status of the yarn.

[0039] Specifically, traditional quality rating methods have two major flaws: First, they "judge quality solely by macroscopic morphology," using macroscopic morphology data as the only basis, leading to the underestimation of potential microscopic morphological defects caused by static electricity (such as fiber arrangement disorder caused by static electricity). Second, the rating dimensions are singular, judging quality only through "pass / fail" or simple scores, failing to distinguish defect types (such as the difference between fuzz and broken yarn) and severity (such as the difference between the impact of slight fuzz and large-area fuzz), resulting in a lack of targeted production control. This step, based on a yarn static state dataset and combined with real-time yarn morphology data (diameter, twist, fuzz, etc.) collected by image recognition equipment, uses correlation analysis to uncover static characteristics strongly correlated with quality defects (such as the static distribution pattern when fuzz exceeds the standard). Based on the identified defect types (such as uneven yarn, broken yarn) and severity (such as defect proportion, impact range), a dynamic rating is generated from both "type" and "severity" dimensions to produce a real-time yarn quality rating. By integrating yarn electrostatic state datasets and morphological data, the strong correlation between "electrostatics and defects" (such as the periodic fluctuations in electrostatic distribution when yarn is uneven) is discovered, enabling defect identification to move from surface observation to mechanism analysis and significantly improving identification accuracy. At the same time, the dual-dimensional dynamic rating not only clarifies the defect type but also classifies the severity, making the rating results more in line with actual production needs.

[0040] S203. Based on the real-time quality rating of the yarn, perform trend analysis on the change of the real-time quality rating of the yarn to generate the subsequent quality change trend of the yarn.

[0041] The subsequent quality change trend of yarn can refer to the predicted direction and extent of quality changes in the yarn during the subsequent spinning process, based on the current electrostatic state, quality rating, and weighted calibration results.

[0042] Specifically, real-time yarn quality ratings only reflect the current quality status. However, textile production is a continuous and dynamic process, and quality status changes constantly with numerous factors. Focusing solely on the current quality rating makes it impossible to predict future quality trends, potentially leading to measures being taken only when quality has severely declined, resulting in a large number of defective products. This step uses real-time yarn quality ratings and known electrostatic adsorption anomaly trajectories to predict subsequent yarn quality trends, mitigating interference with less sensitive products and generating a quality trend based on this prediction. Executing targeted process control instructions based on these trends allows for timely optimization of process parameters, ensuring product quality stability, reducing defective products, providing a basis for proactive process control, and offering forward-looking risk warnings to production personnel to prevent the escalation of quality problems.

[0043] S204. Based on the subsequent quality change trend of the yarn, execute the corresponding process control instructions and output the quality rating log.

[0044] Process control instructions can refer to commands formulated based on the subsequent quality change trends of yarn, used to adjust the parameters of textile equipment. A quality rating log can refer to a log file that records real-time quality rating results, quality change trends, process control instructions, and other information about the yarn.

[0045] Specifically, traditional textile production suffers from severe lag in quality control, often adjusting processes only after defects are discovered, by which time a certain number of defective products have already been generated, resulting in raw material waste and efficiency losses. Furthermore, quality data is stored in a fragmented manner, lacking a systematic log, making it difficult to trace problems and hindering continuous process optimization. This step, based on predicted quality trends, executes process control instructions in advance, controlling quality problems at their nascent stage, achieving "prevention before the event." Simultaneously, the quality rating log fully records real-time rating results, trend predictions, and control instructions, outputting them to users, forming a closed loop of "data collection-analysis-control-recording." This facilitates tracing the root cause of quality problems in specific batches and provides long-term data support for process parameter optimization, upgrading quality control from passive response to proactive optimization, significantly improving production efficiency and quality stability.

[0046] This solution first collects spatiotemporal charge data, then generates a yarn electrostatic state dataset through multimodal sensor fusion, achieving precise quantification of electrostatic factors. It then integrates electrostatic state and morphological data to extract electrostatic characteristics associated with defects, generating real-time quality results through dual-dimensional dynamic rating, improving the comprehensiveness and accuracy of defect identification. By combining yarn category information and electrostatic correlation weights, it calibrates electrostatic data and predicts quality change trends, enhancing the targeted nature of the analysis. Based on these trends, it executes forward-looking process adjustments and outputs quality rating logs to users, forming a closed-loop management system. By generating an electrostatic state dataset through multimodal sensor fusion, identifying defects and assigning dual-dimensional ratings based on morphological data, and then predicting trends and adjusting accordingly based on category information and electrostatic weights, it can accurately detect and quantify the dynamic details of high-speed running yarn in real time. This allows for timely intervention in quality fluctuations, reduces the continuous production of defective products, avoids further waste of raw materials, ultimately enhances product competitiveness and reduces production costs.

[0047] In some embodiments, a spatially deployed ring array of electrostatic sensors is used to collect spatiotemporal charge datasets on key points of the yarn in real time without contact. The spatiotemporal charge dataset includes real-time charge intensity values ​​and real-time charge distribution characteristics. A high frame rate vision sensor is used to capture yarn morphology data and motion trajectory in real time. The spatiotemporal correlation between real-time charge intensity values, real-time charge distribution characteristics, and yarn morphology data is analyzed simultaneously to identify abnormal electrostatic adsorption trajectories caused by quality defects. Based on the spatiotemporal charge dataset, yarn morphology data, and abnormal electrostatic adsorption trajectories, a yarn electrostatic state dataset is generated.

[0048] An electrostatic sensor ring array is a sensing device formed by deploying multiple electrostatic induction rings at preset spatial intervals. It is used to collect charge information on the surface of yarn in a non-contact manner. The real-time charge intensity value can be the charge strength at a specific location on the yarn at a certain moment, reflecting the degree of charge on the yarn surface.

[0049] Real-time charge distribution characteristics can be the spatial distribution pattern of charge on the yarn surface (such as uniform distribution, local aggregation, etc.), reflecting the differences in charge distribution in different regions of the yarn.

[0050] High frame rate vision sensors are image acquisition devices with high shooting frame rates (such as hundreds of frames per second or more), which can capture subtle morphological changes and motion trajectories of yarn during rapid movement.

[0051] Yarn morphology data can be yarn appearance feature data collected by a high frame rate vision sensor, including morphological information such as yarn diameter, number of hairs, and whether there are knots or breaks.

[0052] Abnormal electrostatic adsorption trajectories can be deviations from normal movement patterns caused by yarn quality defects (such as excessive fuzz or uneven fiber distribution). For example, excess fibers at quality defects may produce unexpected swaying or sticking trajectories due to electrostatic adsorption.

[0053] The yarn electrostatic state dataset can be a comprehensive dataset formed by integrating spatiotemporal charge datasets, yarn morphology data, and electrostatic adsorption anomaly trajectories, which is used to comprehensively reflect the state of yarn under the correlation of electrostatic properties and morphological characteristics.

[0054] Specifically, in the yarn spinning process, a single sensing method is insufficient to fully reflect the essence of quality defects: for example, relying solely on electrostatic sensing can detect abnormal charges on the yarn surface (such as uneven charge distribution caused by fiber friction), but it cannot distinguish whether the abnormality is caused by defects in the fiber itself (such as deviations in the blending ratio) or by the external environment (such as changes in humidity), which can easily lead to misjudgments; relying solely on visual sensing can detect macroscopic morphological defects in the yarn (such as broken ends and knots), but it is not sensitive to microscopic defects (such as excessive internal fuzz) and is greatly affected by changes in light and yarn color; yarn quality defects often manifest simultaneously as abnormal physical morphology and imbalanced charge distribution (such as loose fibers accumulating excessive charge due to friction, leading to the adsorption of impurities and causing trajectory deviation), and without spatiotemporal correlation analysis, key diagnostic information will be missed. To address the above issues, this step first utilizes a spatially deployed ring array of electrostatic sensors to non-contactly sense the surface charge of the yarn via ring electrodes. This charge is then converted into a voltage signal by a charge amplifier and digitized by an ADC module to generate a spatiotemporal charge dataset. Simultaneously, a high-frame-rate visual sensor captures yarn motion images under polarized light illumination. An FPGA executes a real-time yarn morphology algorithm to output yarn shape data (e.g., diameter 0.12mm). Based on the displacement changes of yarn feature points in multiple consecutive frames, a 3D motion reconstruction algorithm generates the yarn trajectory (e.g., linear velocity > 2mm / s). Next, the central controller receives these data streams and, according to a precise clock, synchronizes them. The system aligns timestamps and slices them into time windows (e.g., 10ms) for spatiotemporal correlation analysis. When a sudden change in charge intensity is detected (e.g., ΔQ / Δt > 5pC / ms), the vision system is triggered to focus on the corresponding position. The angle similarity between the charge gradient vector direction and the yarn movement offset direction is calculated. If the angle difference is small (e.g., <10°) and the offset acceleration is large (e.g., >0.5m / s²), it is determined to be an abnormal electrostatic adsorption trajectory. Finally, a yarn electrostatic state dataset is generated based on the analysis results. This dataset contains charge distribution characteristics, yarn morphology data, and abnormal trajectory information, and is uploaded to the manufacturing execution system through a dynamic update mechanism (e.g., generating one record every 100ms).

[0055] This solution employs a multimodal sensing fusion strategy to enable electrostatic sensing to capture microscopic charge changes and visual sensing to record macroscopic morphological features. The combination of the two can cover all dimensions of quality information, from inter-fiber interactions (microscopic) to the overall yarn structure (macroscopic). The abnormal trajectory of electrostatic adsorption correlates charge changes with motion state, which can not only identify defects but also trace their causes (such as trajectory deviation accompanied by charge accumulation, which may originate from local unevenness), providing a precise basis for subsequent process control.

[0056] In some embodiments, based on the yarn electrostatic state dataset and real-time acquired yarn morphology data, yarn defect types are identified, and the severity of the corresponding explicit defects is determined. Yarn defect types include excessive hairiness, uneven yarn count, and fiber knots. Several yarn defect types are normalized to a unified numerical range to determine the normalized yarn defect types. Based on the normalized yarn defect types and the severity of explicit defects, a dynamic rating score is determined. Based on the dynamic rating score, a real-time yarn quality rating is generated, which is divided into three levels: excellent, medium, and poor.

[0057] Yarn defect types can refer to three typical quality anomalies that occur during yarn production: excessive hairiness (fibers protruding from the yarn surface), uneven yarn (yarn diameter fluctuations exceeding limits), and fiber knots (foreign matter mixed in or fibers entangled). The severity of visible defects refers to the intuitive severity of a particular yarn defect type under the current production conditions, usually described by quantitative indicators. Real-time yarn quality rating refers to the yarn quality grade divided according to dynamic rating scores, categorized into three levels: "Excellent," "Medium," and "Poor."

[0058] Specifically, traditional yarn quality inspection often relies on single visual features (such as observing shape only through a camera), easily overlooking abnormal electrostatic properties caused by changes in fiber material and friction (e.g., excessive hairiness can lead to localized charge accumulation, and uneven yarn can cause charge distribution fluctuations). Different types of yarn defects have significantly different impacts on the performance of the final product: for example, excessive hairiness mainly affects the appearance of the fabric, uneven yarn directly reduces yarn strength, and fiber knots may cause yarn breakage during weaving. If only the type of defect is identified without distinguishing its severity, it is impossible to provide a precise basis for process adjustment (e.g., slight hairiness can be ignored, while severe hairiness requires immediate adjustment of drafting parameters). The original evaluation indicators for yarn defects have significant dimensional differences: the evaluation indicator for excessive hairiness is "number of hairs per unit length" (integer), uneven yarn is "coefficient of variation in diameter" (percentage), and fiber knots are "knot density" (numbers / meter). If a comprehensive evaluation is made directly based on the original indicators, the results will be distorted due to differences in numerical range and physical meaning (e.g., "10 hairs / meter" and "5% diameter variation" cannot be directly compared). To address the above issues, this step first uses a high frame rate visual sensor (e.g., 2000fps) to capture the yarn surface morphology in real time, simultaneously combining this with a charge distribution heatmap generated by a ring array of electrostatic sensors to accurately locate three types of defect areas: excessive fuzz, uneven yarn, and fiber knots. Then, for the fuzzy areas, the number of fuzzes per unit length (e.g., 15 fibers / cm) and average charge intensity (e.g., 0.5nC) are calculated; for uneven yarn sections, the diameter fluctuation variance (e.g., 0.2mm²) is extracted; and for fiber knots, the charge peak value (e.g., 8nC) and impurity geometry (e.g., 0.3mm²) are recorded. Next, the historical extreme value database for the current production batch (e.g., ...) is retrieved. The extreme range of feather density is 0-85 feathers / cm. The original defect value is converted into a uniform range (e.g., [0,1]) using the range normalization formula. Then, based on the real-time determination of defect severity (e.g., a coefficient of 0.5 for moderate feather defects), the normalized value is multiplied by the corresponding weight (e.g., feather weight 0.3) and the severity coefficient and accumulated (e.g., 0.18×0.3×0.5≈0.027) to obtain the dynamic rating total score (e.g., 0.45). Finally, the scoring results are mapped to a two-dimensional rating: a score ≤0.3 outputs an excellent rating, a score between 0.3 and 0.7 (e.g., 0.45) outputs a medium rating, and a score >0.7 outputs a poor rating.

[0059] This solution integrates electrostatic state data and morphological data to identify not only physical defects in yarn (such as fuzz and unevenness) but also latent defects (such as abnormal charges caused by fiber knots) through changes in electrostatic properties. This avoids the limitations of a single data source and improves the comprehensiveness and accuracy of quality defect identification. The dynamic rating score is updated in real time with the production process, reflecting quality fluctuations promptly and better reflecting actual production conditions compared to traditional static sampling rating. Normalization eliminates evaluation differences between different defects, and weighted calculation makes the rating results more reflective of the comprehensive impact of defects on quality, providing a precise basis for process adjustments.

[0060] In some embodiments, a yarn appearance image and corresponding real-time charge intensity value are acquired in real time. Based on the yarn appearance image, the following hair density gridding analysis steps are performed: when the number of discrete fibers in several consecutive detection data exceeds a preset first density threshold and the average electrostatic intensity is greater than a preset multiple of the visual hair number, it is determined to be a local hair aggregation defect; when the number of discrete fibers in a unit detection area exceeding a preset proportion exceeds a preset second density threshold and the overall electrostatic distribution non-uniformity is greater than a preset diffuse hair proportion, it is determined to be a diffuse hair excess defect; based on the combination pattern of the number of local aggregation areas and the diffuse distribution proportion, combined with the coupling coefficient between the electrostatic intensity increase and hair density, the spatial distribution state of hair is analyzed, and the hair density manifestation degree is quantified.

[0061] Yarn appearance images can be images of the yarn surface captured by high frame rate vision sensors. They can intuitively present the appearance characteristics of the yarn, such as the distribution of hairiness and the arrangement of fibers, and are a direct basis for visual analysis of hairiness defects.

[0062] The real-time charge intensity value can be data collected from key points of the yarn by a spatially deployed ring array of electrostatic sensors. It reflects the strength of static electricity generated by yarn friction and is strongly correlated with the hair density. The preset first density threshold can be a pre-set critical value for the number of discrete fibers used to determine whether hair in a local area is "aggregated". It is obtained by statistical analysis of historical high-quality yarn data to ensure a consistent standard for judging local aggregation defects.

[0063] Localized fuzz aggregation defects refer to a sudden surge in the number of discrete fibers and a significantly higher electrostatic intensity within a localized area of ​​the yarn surface. This is typically caused by uneven fiber combing and significantly impacts the risk of "thread jamming" in subsequent weaving. The preset second density threshold is a pre-defined critical value for the number of discrete fibers within a unit detection area to determine whether fuzz exceeds the standard. Its value is usually lower than the preset first density threshold (because the number of fibers in a single area of ​​a diffuse defect is lower than that of a localized aggregation, but the distribution range is wider). Excessive diffuse fuzz defects refer to a defect where the number of discrete fibers in multiple areas (exceeding a preset proportion) on the yarn surface exceeds the standard, and the overall electrostatic distribution is chaotic. This is usually caused by poor fiber raw material quality or unstable drafting processes and significantly affects yarn strength and dyeing uniformity.

[0064] The spatial distribution of hair refers to the spatial characteristics of hair distribution on the yarn surface, such as its location, range, and density, including "local concentration" and "global dispersion," which are key criteria for distinguishing defect types.

[0065] The degree of hair density visibility can be comprehensively characterized by quantitative indicators (such as the number of local clustered areas, the proportion of diffuse distribution, coupling coefficient, etc.) to determine the obviousness of hair defects in appearance and electrostatic properties. It is the core basis for judging the severity of visible defects.

[0066] Specifically, yarn hairiness is one of the core indicators of yarn quality. Its quantity and distribution directly affect the stability of subsequent weaving processes (such as breakage rate and fabric smoothness) and dyeing processes (such as evenness), as well as the appearance and performance of the final product. Traditional yarn hairiness detection methods mainly rely on visual image analysis, which has two major limitations: First, they are easily affected by lighting and image noise, leading to "misjudgment" (such as misjudging dust as hairiness). Second, they judge the severity solely by the quantity of hairiness, ignoring the distribution characteristics of hairiness (the impact of local aggregation and global dispersion on the process is significantly different), resulting in a lack of targeted subsequent process control. The causes and effects of local hairiness aggregation (such as a sudden appearance of a large amount of hairiness in a 1mm area) and diffuse hairiness excess (such as a small amount of hairiness throughout the yarn but exceeding the standard overall) are completely different. If they are not distinguished and the same process control strategy is applied uniformly, it will lead to "treating the symptoms but not the root cause". To address the above issues, this step first performs defect determination: by dividing the yarn appearance image into a grid, the number of discrete fibers in each unit detection area (e.g., 1mm × 1mm) is counted; for localized fuzz aggregation defects, when the number of fibers in a certain area exceeds a preset first density threshold (e.g., 50 fibers / mm²) in several consecutive detections (e.g., 5 consecutive detections) and the average electrostatic intensity of that area is greater than a preset multiple (e.g., 1.2 times) of the visual fuzz count, it is marked as an aggregation defect area and the number is recorded; for diffuse fuzz excess defects, the number of fibers in all unit areas exceeding a preset second density threshold is calculated. The proportion of values ​​(e.g., 30 hairs / mm²) is determined to be a dispersion defect if the excess proportion exceeds the preset proportion (e.g., 20%) and the overall electrostatic distribution unevenness is greater than the preset dispersion hair proportion (e.g., 0.4). Then, the spatial distribution state of the hairs is analyzed: it is classified according to the combination pattern of the number of local clustered areas and the dispersion distribution proportion (e.g., pattern A is high clustered number + low dispersion proportion, pattern B is the opposite), and the degree of dominance is quantified based on the coupling coefficient (e.g., electrostatic intensity increase / hair density increase). The specific formula is: degree of dominance = (number of clustered areas × 0.6 + dispersion proportion × 0.4) × coupling coefficient.

[0067] This solution utilizes a dual-dimensional analysis of "visual images + electrostatic properties" to effectively improve the accuracy of yarn hair defect identification and reduce misjudgments caused by other external factors. Simultaneously, it enables refined classification of yarn hair defects, accurately distinguishing between localized hair aggregation and diffuse excessive hair, providing a clear direction for subsequent process control and avoiding blind adjustments. By comprehensively considering indicators such as the number of localized aggregation areas, the coupling coefficient between electrostatic intensity increase and hair density, the quantification of the severity of visible defects becomes more scientific, providing precise data support for quality rating. This enhances the controllability of the production process, helps reduce problems such as weaving breakage and uneven dyeing caused by yarn hair defects, and improves the stability and consistency of yarn quality.

[0068] In some embodiments, a yarn axial image sequence is acquired in real time. Based on the yarn axial image sequence, the following three-dimensional yarn evenness fluctuation analysis steps are performed: Diameter anomaly regions where the diameter difference between adjacent sampling points exceeds the diameter tolerance threshold are marked as abrupt non-uniform regions. Simultaneously, the rate of change of the electrostatic adsorption trajectory direction angle of the abnormal diameter region is detected. When the rate of change of the direction angle exceeds a preset turbulence threshold, the current diameter anomaly region is marked as electrostatic turbulence abrupt non-uniformity. Trend anomaly regions corresponding to a unidirectional continuous increase / decrease in diameter exceeding a preset number of points are marked as gradual non-uniform regions. Simultaneously, the slope of the electrostatic gradient distribution in the gradual region is acquired. When the slope is in the same direction as the diameter change trend and exceeds the associated threshold, it is marked as electrostatic accumulation gradual non-uniformity. The proportions of the lengths of abrupt non-uniform regions and gradual non-uniform regions in the unit detection area are determined. Combined with the maximum diameter deviation value positively correlated with the degree of yarn evenness and the electrostatic turbulence intensity parameter, and the electrostatic gradient consistency parameter negatively correlated with the degree of yarn evenness, the degree of yarn evenness is quantified.

[0069] A yarn axial image sequence can be a collection of axial profile images of the yarn surface continuously captured along the yarn's direction of movement. These images are acquired at fixed time intervals using a high-frame-rate vision sensor, forming time-series image data reflecting the continuous change in yarn diameter. Three-dimensional yarn evenness analysis refers to the analysis of the three-dimensional spatial fluctuation characteristics of diameter changes along the yarn's axial direction, combined with changes in electrostatic parameters, to identify the type and distribution pattern of yarn unevenness. A diameter tolerance threshold refers to a preset maximum allowable difference in diameter between adjacent sampling points of the yarn; exceeding this value is considered a diameter anomaly.

[0070] Abrupt uneven regions can refer to diameter abnormal regions where the diameter difference between adjacent sampling points along the yarn axis exceeds the diameter tolerance threshold, manifested as a significant abrupt change in diameter over a short distance.

[0071] The rate of change of the direction angle of the electrostatic adsorption trajectory can refer to the degree of change of the direction angle of the trajectory formed by the electrostatic adsorption of surrounding particles on the yarn surface per unit time, reflecting the stability of the electrostatic field.

[0072] Electrostatic turbulence-type abrupt non-uniformity can refer to a defect type in which the rate of change of the direction angle of electrostatic adsorption trajectory exceeds a preset turbulence threshold in the abrupt non-uniform region, indicating a strong correlation between diameter abrupt change and electrostatic field disturbance.

[0073] Gradual uneven regions can refer to abnormal trend regions corresponding to a unidirectional continuous increase / decrease in diameter that exceeds a preset number of points along the yarn axis, manifested as a slow change in diameter over a long distance.

[0074] Electrostatic accumulation-type gradual unevenness refers to a defect type in which the slope of the electrostatic gradient distribution in the gradually uneven region is in the same direction as the diameter change trend and exceeds the correlation threshold, indicating a strong correlation between diameter gradualness and electrostatic accumulation / dissipation. The maximum diameter deviation value refers to the maximum difference between the yarn diameter and the standard diameter within a unit of inspection area, directly reflecting the severity of yarn unevenness.

[0075] Electrostatic disturbance intensity parameter refers to a quantitative index used to characterize the degree of electrostatic field disturbance within a unit detection area, which integrates characteristics such as the rate of change of the direction angle of electrostatic adsorption trajectory and the fluctuation of electrostatic gradient.

[0076] The electrostatic gradient uniformity parameter can be a quantitative index used to characterize the uniformity of the electrostatic gradient distribution within a unit detection area. The higher the uniformity, the more stable the electrostatic field.

[0077] The degree of unevenness can refer to the extent to which unevenness defects are manifested in yarn morphology and electrostatic characteristics. It is a defect severity index quantified by a combination of multiple parameters.

[0078] Specifically, traditional methods distinguish between abrupt and gradual defects solely based on diameter change characteristics. However, in actual production, some abrupt diameter changes may be accidental fluctuations caused by short-term mechanical vibrations rather than genuine quality defects. Some gradual trends may be "false diameter changes" caused by electrostatic adsorption of fibers. If only morphological data is relied upon, normal fluctuations are easily misjudged as defects, or latent defects related to electrostatics may be overlooked, leading to a high rate of misjudgment of defect types. Moreover, traditional methods only measure the severity of unevenness by diameter deviation values, but ignore the amplification effect of electrostatic characteristics on defects. For example, under the same diameter deviation, if accompanied by strong electrostatic disturbances, it may lead to further fiber aggregation during subsequent weaving, and the defect will continue to worsen. However, if the electrostatic field is stable, the defect may remain stable. Relying solely on the quantitative results of diameter parameters cannot reflect the "potential risk of deterioration" of defects, resulting in insufficient accuracy in quality rating. To address the above issues, this step first uses a high frame rate visual sensor to capture an axial image sequence along the yarn's travel direction at a fixed frame rate (e.g., 500 frames / second). Simultaneously, a circular array of electrostatic sensors collects electrostatic adsorption trajectory data for the corresponding segments. Then, the yarn diameter is extracted frame by frame from the axial image sequence, and the diameter difference between adjacent sampling points is calculated. When the difference exceeds a preset diameter tolerance threshold (e.g., ±8% of the nominal diameter), it is marked as a candidate abrupt change region. Simultaneously, the rate of change of the electrostatic adsorption trajectory direction angle in this region is analyzed—if the rate of change exceeds a preset turbulence threshold (e.g., ≥15° / ms), it is determined to be an electrostatic turbulence-type abrupt change; otherwise, it is marked as a normal abrupt change region. Simultaneously, the image is detected... Regions in the sequence with a continuous unidirectional diameter increase or decrease trend exceeding a preset number of points (e.g., ≥5 sampling points) are marked as candidate gradient regions, and their electrostatic gradient distribution slope is calculated. When the slope is in the same direction as the diameter change trend (e.g., slope > 0 when diameter increases) and the absolute value exceeds the correlation threshold (e.g., ≥ 0.3), it is judged as electrostatic accumulation type gradient non-uniformity. Finally, the proportion of abrupt and gradient regions within a unit detection length (e.g., 1 meter) is statistically analyzed, and positive correlation parameters such as the maximum diameter deviation value and electrostatic turbulence intensity (e.g., the mean of the rate of change of direction angle) and negative correlation parameters such as electrostatic gradient consistency (e.g., the inverse of the variance of the slope of adjacent regions) are integrated to finally quantify the degree of strip non-uniformity ranging from 0 to 1.

[0079] This solution integrates yarn morphology data and electrostatic characteristics to accurately distinguish between abrupt and gradual unevenness, identify whether defects are dominated by electrostatics, significantly reduce the misjudgment rate, and improve the alignment between defect type identification and actual production causes. By comprehensively quantifying the degree of unevenness manifestation using multi-dimensional parameters, it reflects both the current severity and the potential risk of deterioration, making the results more comprehensive and predictive. After clarifying the type and cause of unevenness defects, subsequent process control can be adjusted in a targeted manner, improving the efficiency of quality improvement.

[0080] In some embodiments, real-time scanning imaging data of the yarn surface is acquired, and the corresponding electrostatic intensity extreme values ​​are captured simultaneously. The following impurity multi-scale feature extraction steps are performed to clarify the impurity attributes: when there is a dark clump structure with irregular boundaries and a charge decay rate less than a preset decay rate threshold, the dark clump structure is determined to be a fiber agglomerate; when there is a reflective particle with geometric edges and an electrostatic intensity extreme value greater than a preset multiple of the particle size, the reflective particle is determined to be a hard impurity; when there is a radial fiber entanglement structure with an oscillating electrostatic intensity, the radial fiber entanglement structure is determined to be a yarn defect nodule; based on the size level, optical contrast, and boundary sharpness of the impurity multi-scale features, the size level, optical contrast, and boundary sharpness are weighted according to the impurity attributes to quantify the degree of fiber agglomeration.

[0081] Yarn surface scanning imaging data can refer to image sequence data containing surface details (such as texture, foreign objects, structural anomalies, etc.) obtained by continuously scanning the yarn surface using a high frame rate vision sensor.

[0082] The extreme value of electrostatic intensity can refer to the maximum or minimum value (i.e., the extreme value of electrostatic intensity) that appears among the electrostatic intensity values ​​at different locations on the yarn surface that are collected simultaneously during the scanning process of the yarn surface.

[0083] The charge decay rate can be the rate of change of the static charge intensity on the yarn surface from its peak value to a stable value. Fiber agglomerates refer to irregular clump-like structures formed by fibers entanglement and adhesion during yarn production. Their material is the same as the yarn itself (mostly fibers), and they have strong fiber aggregation characteristics.

[0084] Hard impurities can refer to non-fibrous hard foreign objects (such as metal scraps, plastic particles, sand, etc.) mixed in yarn, which have clear geometric edges and high hardness.

[0085] Yarn defects and knots can refer to radial entanglement structures formed by abnormal fiber arrangement during the twisting or drafting process of yarn. They manifest as localized loose fibers that are entangled into knots, and are a structural defect of the yarn itself.

[0086] Size rating refers to a grading standard based on the physical dimensions of impurities (such as length, diameter, area, etc.), used to characterize the degree to which the physical size of impurities affects yarn quality. Optical contrast refers to the difference in optical signals (such as grayscale value, brightness, color saturation, etc.) between the impurity area and the yarn body area in the scanned imaging data. The greater the difference, the easier it is for the impurity to be visually identified. Boundary sharpness refers to the clarity of the boundary between the impurity and the yarn body, determined by parameters such as boundary continuity and sharpness calculated using image edge detection algorithms. The sharper the boundary, the higher the distinguishability between the impurity and the yarn body.

[0087] The degree of fiber knot and impurity manifestation can refer to the degree to which impurities are obvious on the yarn surface. It is a quantitative characterization of the impact of impurities on yarn quality (the higher the value, the more significant the negative impact of impurities on yarn quality).

[0088] Specifically, in yarn textile quality rating, impurities are one of the key factors affecting yarn performance (such as strength, uniformity, and subsequent weaving stability). However, the impact mechanisms of different types of impurities on quality vary significantly: fiber clumps can cause uneven yarn thickness in certain areas, which can easily lead to yarn breakage or pilling during weaving; hard impurities may wear down textile equipment (such as the reed and rollers of the loom), shortening equipment life, and may also form holes on the fabric surface; yarn defects and knots can disrupt the continuity of the yarn, leading to uneven dyeing or protrusions on the fabric surface, affecting the appearance quality. If "impurities" are simply used as a general indicator of quality defects without distinguishing their attributes or quantifying their impact, the quality rating results will be distorted. For example, for fiber clumps and hard impurities of the same size, the latter is far more harmful to production than the former. If they are treated equally, the process control instructions will lose their specificity (e.g., hard impurities require emergency shutdown for cleaning, while fiber clumps can be alleviated by adjusting the drafting parameters). To address the above issues, this step involves the system simultaneously triggering a high-frame-rate visual sensor and an electrostatic sensor ring array to acquire real-time scanning imaging data of the yarn surface and the corresponding extreme electrostatic intensity values ​​at a high frequency (e.g., 2kHz). Subsequently, multi-scale feature extraction of impurities is performed to clarify their attributes: when a dark, clump-like structure with irregular boundaries is located in the imaging data, and the charge decay rate of its corresponding electrostatic signal remains below a preset threshold (e.g., 50V / s) for several frames (e.g., 3 frames), it is determined to be a fiber agglomerate; when a reflective particle region with geometrical edges is detected, and the ratio of its extreme electrostatic intensity value to its physical size is greater than a preset multiple (e.g., 8 times for metallic impurities or 5 times for gravel impurities), it is determined to be a hard impurity; when a radial fiber entanglement structure is identified, and its corresponding electrostatic signal decays within a short time (e.g., 0.5V / s), it is determined to be a hard impurity. When a certain number of oscillations (e.g., ≥3 times) occur within 5 seconds (amplitude change rate >20%), it is judged as a yarn defect nodule. Then, based on the identified impurity attributes, its size grade (e.g., micro=0.1, small=0.3, medium=0.6, large=1.0), optical contrast (normalized value), and boundary sharpness (normalized gradient value) are extracted respectively. Then, according to the preset weights corresponding to the attributes (e.g., nodule: size weight 0.5, contrast weight 0.3, boundary weight 0.2; hard impurity: size weight 0.4, contrast weight 0.4, boundary weight 0.2; nodule: size weight 0.6, contrast weight 0.2, boundary weight 0.2), a weighted sum is performed to finally quantify the fiber nodule visibility value in the range of 0 to 1.

[0089] This solution integrates yarn surface scanning imaging data with extreme electrostatic intensity values, combined with multi-scale feature extraction, to effectively distinguish fiber clumps, hard impurities, and yarn defects, avoiding missed detections or misjudgments caused by single visual inspection and improving the accuracy of impurity identification. By weighting size grade, optical contrast, and boundary sharpness, the impact of different impurities on yarn quality can be scientifically quantified, making the quality rating results more in line with actual production needs and achieving precise quantification of quality impact. After clarifying the impurity attributes and dominance, targeted process control instructions can be generated, supporting precise process control, thereby reducing ineffective downtime or over-control caused by impurity misjudgment, lowering production costs, reducing wear and tear on equipment caused by hard impurities, extending the service life of textile equipment, and improving production efficiency.

[0090] In some embodiments, based on a preset rule table, identified yarn defect types are mapped to defect type identifiers, and assessed severity of visible defects is mapped to severity level identifiers. Several defect type identifiers are concatenated with their corresponding severity level identifiers, and the concatenated result is used as a unit rating code. After all unit rating codes are concatenated, a second concatenation is performed on the unit rating codes to generate a composite rating code. According to preset mapping rules, the composite rating code is mapped to an initial quality level of excellent / medium / poor. The mapping rules include: determining the total number of all defects as the defect density, and identifying the highest severity level identifier in the composite rating code; if the defect density is greater than a first density threshold or the highest severity level identifier reaches the first severity level threshold, it is judged as poor; if the defect density does not meet the aforementioned condition for judging as poor but is greater than a second density threshold or the highest severity level identifier reaches the second severity level threshold, it is judged as medium; if the defect density does not exceed the second density threshold and the highest severity level identifier does not reach the second severity level threshold, it is judged as excellent; wherein, the first density threshold is greater than the second density threshold, and the first severity level threshold is higher than the second severity level threshold.

[0091] The preset rule table can be a pre-established mapping table used to convert yarn defect types and severity into standardized identifiers. The defect type identifier can be a unique code representing the category of yarn defect (e.g., "B" indicates excessive hairiness, "C" indicates uneven yarn). The severity level identifier can be a hierarchical code characterizing the severity of the defect (e.g., "1" indicates minor, "2" indicates severe).

[0092] The unit rating code can be a complete rating identifier for a single defect, which is composed of a defect type identifier and a severity level identifier (e.g., "C1 indicates" slight unevenness).

[0093] A composite rating code can be a global identifier integrating all unit rating codes (e.g., "B3+C1" indicates severe excessive hairiness and slight unevenness). The first density threshold can be the critical value for the number of defects to determine "poor" yarn quality; it is a preset standard for a higher number of defects. The second density threshold can be the critical value for the number of defects to determine "medium" yarn quality, and it is less than the first density threshold; it is a preset standard for a medium number of defects. The first severity level threshold can be the critical value for the severity of yarn quality to determine "poor" quality; it is a preset standard for a higher severity level. The second severity level threshold can be the critical value for the severity of yarn quality to determine "medium" quality, and it is lower than the first severity level threshold.

[0094] Specifically, quality rating is a core link connecting quality inspection and process control, directly affecting production efficiency, product qualification rate, and production cost. Traditional yarn quality rating methods have limitations: First, they suffer from the one-sidedness of assessing a single defect. Traditional methods often rate a single defect (such as only assessing the amount of hairiness), ignoring the comprehensive impact of multiple coexisting defects on overall quality. For example, a yarn may have both slightly excessive hairiness and moderate unevenness at the same time, and a single defect rating cannot reflect its true quality level. Second, the grading is unclear. The lack of clearly defined thresholds for "excellent," "medium," and "poor" grades makes it difficult for production personnel to accurately determine whether the yarn meets production requirements, thus delaying process adjustments and increasing the defect rate. To address the above issues, this step first calls a preset rule table to map the identified defect types to standardized identifiers (e.g., excessive fuzz is mapped to "B"). Simultaneously, a severity level identifier (e.g., "3") is generated based on the severity of the explicit defect (e.g., fuzz density reaching level 3). Then, a string concatenation operation is performed on each defect to generate a unit rating code containing type and level information (e.g., "B3"). Next, all defects are traversed, and the unit rating codes are concatenated again according to the detection sequence to form a global composite rating code (e.g., "B2+C1", representing two types of defects: excessive fuzz (level 2) and uneven stripe (level 1)). Finally, the composite rating code is statistically analyzed. The total number of defect type identifiers in the rating code is used as the defect density (e.g., 2 defects / meter), and the highest value (2 in this example) of all severity level identifiers is extracted. Finally, the quality level is determined based on a dual threshold mechanism: if the defect density exceeds the first density threshold (e.g., 5 defects / meter) or the highest severity level reaches the first severity level threshold (e.g., 3), it is judged as "poor"; if it does not reach "poor" but the defect density exceeds the second density threshold (e.g., 2 defects / meter) or the highest severity level reaches the second severity level threshold (e.g., 2), it is judged as "medium"; only when the defect density is ≤2 defects / meter and the highest severity level is lower than "2" is it judged as "excellent".

[0095] This solution integrates all defect information using composite rating codes, avoiding the limitations of single-defect assessments and accurately reflecting the overall quality of the yarn. It clearly defines quality grade boundaries, enabling production personnel to quickly identify yarn quality status and take timely corresponding measures, reducing production losses caused by ambiguous quality judgments. It supports real-time process control, providing a clear basis for process optimization, helping to reduce the defect rate and improve the stability of yarn product quality and production efficiency. Simultaneously, it promotes standardized management of quality data; standardized coding methods such as composite rating codes facilitate the storage, transmission, and analysis of quality data, laying the foundation for subsequent quality traceability, production optimization, and big data analysis.

[0096] In some embodiments, based on the abnormal electrostatic adsorption trajectory, trajectory temporal dynamic features including trajectory fluctuation frequency, directional consistency, and intensity variation amplitude are extracted; trajectory evolution patterns are identified based on the dynamic features: when the trajectory fluctuation frequency is lower than a preset low-frequency threshold and the directional consistency is higher than a stable threshold, it is a stable and gradual trend; when the trajectory fluctuation frequency is higher than a preset high-frequency threshold and the intensity variation amplitude continues to increase, it is a deterioration and diffusion pattern; when the directional consistency is lower than a disorder threshold and the intensity variation amplitude decreases periodically, it is an intermittent oscillation pattern; the trajectory evolution pattern is mapped to a quality trend: the stable and gradual trend is mapped to a defect mitigation trend, the deterioration and diffusion pattern is mapped to a defect escalation trend, and the intermittent oscillation pattern is mapped to a quality fluctuation trend; based on the quality trend and the severity of the explicit defects, the subsequent quality change trend of the yarn is generated.

[0097] An abnormal electrostatic adsorption trajectory can refer to a path where the electrostatic adsorption trajectory deviates from the normal state due to yarn quality defects. The trajectory's temporal dynamic characteristics refer to a set of features describing the changing patterns of the abnormal electrostatic adsorption trajectory over time. The trajectory fluctuation frequency is the number of fluctuations in the abnormal electrostatic adsorption trajectory per unit time, reflecting the frequency of trajectory changes. Directional consistency refers to the degree to which the direction of the abnormal electrostatic adsorption trajectory remains consistent during movement; a higher value indicates a more stable trajectory direction. The intensity variation amplitude is the range of change in the electrostatic intensity corresponding to the abnormal electrostatic adsorption trajectory over time, reflecting the magnitude of fluctuations in electrostatic intensity. The trajectory evolution pattern refers to the typical patterns of change in the abnormal electrostatic adsorption trajectory over time, summarized based on the trajectory's temporal dynamic characteristics. A stable and gradual pattern can be an evolutionary state with low trajectory fluctuation frequency and high directional consistency. A deterioration and diffusion pattern can be an evolutionary state with high trajectory fluctuation frequency and a continuously increasing intensity variation amplitude. An intermittent oscillation pattern can be an evolutionary state with low trajectory directional consistency and a periodically decreasing intensity variation amplitude. A defect mitigation trend can be a trend of gradually reducing yarn quality defects. A defect escalation trend can be a trend of gradually increasing yarn quality defects. Quality fluctuation trends can be characterized by repeated and unstable changes in yarn quality defects.

[0098] Specifically, real-time quality ratings only reflect the quality status at a certain moment, while yarn quality is a dynamic process: factors such as wear and tear on textile equipment, fluctuations in raw material properties, and changes in environmental temperature and humidity can all cause yarn defects to evolve from "minor" to "severe," or from "stable" to "fluctuating." If only real-time quality ratings are relied upon for process control, it is often only possible to "passively deal with" defects that have already appeared, and it is impossible to prevent the expansion or deterioration of defects in advance. This may lead to a large number of unqualified products and increase production costs. For example, when the yarn has slight "unevenness," the real-time quality rating may be "medium." However, if its electrostatic adsorption trajectory has shown the characteristics of "increased fluctuation frequency and increased intensity variation" (indicating that the defect will escalate), if not intervened in time, it may develop into severe unevenness in a short period of time, and the rating will drop to "poor." At this point, adjusting the process will have already resulted in a large amount of yarn waste. To address the above issues, this embodiment first acquires abnormal electrostatic adsorption trajectory data, and then extracts key dynamic features from the trajectory's temporal information: The trajectory fluctuation frequency (e.g., 0.5Hz) is calculated by analyzing changes in trajectory direction / velocity; directional consistency (e.g., 0.85, higher values ​​indicate greater consistency) is calculated by statistically analyzing the dispersion of the direction vector within a continuous time window; and the intensity variation amplitude (e.g., 0.3) is determined by calculating the range or standard deviation of the adsorption intensity signal within the window. Based on these quantitative features, combined with preset discrimination thresholds (e.g., low-frequency threshold 0.3Hz, high-frequency threshold 1.2Hz, stable directional consistency threshold 0.8, disordered threshold 0.6) and observing the trend of intensity variation amplitude changes (continuous increase, periodic decrease, or stabilization), the trajectory evolution pattern is identified—if the fluctuation frequency is below the low-frequency threshold (e.g., 0.25Hz) and the directional consistency is above the stable threshold (e.g., 0.85), it is determined to be a stable and gradual pattern; if the fluctuation... If the frequency is higher than the high-frequency threshold (e.g., 1.3Hz) and the intensity variation amplitude continues to increase (e.g., from 0.2 to 0.35), it is determined to be a deterioration diffusion mode; if the directional consistency is lower than the disorder threshold (e.g., 0.55) and the intensity variation amplitude decreases periodically (e.g., the amplitude periodically decreases from 0.4 to 0.1), it is determined to be an intermittent oscillation mode; then the identified modes are directly mapped to the corresponding quality trends (stable slowdown -> defect mitigation trend, deterioration diffusion -> defect escalation trend, intermittent oscillation -> quality fluctuation trend); finally, combined with the severity of the current explicit defects assessed in the aforementioned embodiments (e.g., the severity of hairiness defect is 0.7, and the severity of uneven yarn defect is 0.4), a comprehensive trend of subsequent yarn quality changes is generated (e.g., "Prediction: Defect escalation trend (based on deterioration diffusion mode), the current severity of hairiness defect is high (0.7), and immediate intervention is required"), providing a decision-making basis for process control that includes the future predicted direction and the current severity.

[0099] This solution utilizes the analysis of trajectory evolution patterns combined with the severity of explicit defects to correct quality trends. It can predict the changing trends of yarn quality defects before they significantly worsen, and adjust strategies to better suit actual production conditions, buying time for process adjustments and reducing the generation of defective products. By analyzing the changing trends of yarn quality, it combines "real-time rating" with "future prediction," achieving a shift from "passive response" to "proactive prevention." This is irreplaceable and necessary for improving the stability of yarn textile quality, reducing production costs, and increasing production efficiency.

[0100] In some embodiments, differentiated control is performed based on the type of subsequent quality change trend of the yarn: if it is determined to be a defect escalation trend, linkage control is triggered to reduce spinning speed and increase environmental humidification; if it is determined to be a defect mitigation trend, the current process is maintained and the detection density is reduced; if it is determined to be a quality fluctuation trend, the environmental temperature and humidity are dynamically adjusted and static electricity neutralization is initiated in a targeted manner; the real-time quality rating of the yarn, the subsequent quality change trend of the yarn, and the control instructions are associated with the production line batch and time to generate and output a quality rating log.

[0101] The type of quality change trend can refer to the specific direction of subsequent quality changes in yarn, including defect escalation trend, defect mitigation trend, and quality fluctuation trend.

[0102] A quality rating log can be a document that records the real-time quality rating of yarn, quality change trends, control instructions, production line batches and time, and is used for quality traceability and process optimization.

[0103] Specifically, in the yarn spinning process, quality control is not simply a closed loop of "inspection-rating." It requires proactive regulation through anticipating quality change trends, and continuous optimization supported by complete log records. A lack of complete logs leads to untraceable quality issues and difficulty in accumulating regulatory experience, hindering continuous production line improvement. To address these issues, this step receives real-time data on subsequent yarn quality change trends and parses its trend classification tags (escalation / mitigation / fluctuation). Then, it matches the data to a preset regulatory rule library: if the trend is determined to be escalating, a speed reduction command is sent to the spinning equipment (e.g., from 300m / min to 250m / min) and the humidification system is simultaneously triggered to increase the ambient humidity (e.g., from 50% to 65%); if the trend is determined to be mitigating, the current process parameters (e.g., spinning speed 300m / min, humidity 50%) are maintained, and the sampling frequency of the detection sensors is reduced. Rate (e.g., decreasing from 100Hz to 50Hz); if determined to be a quality fluctuation trend, dynamically adjust the ambient temperature and humidity (e.g., fine-tune within ±2℃ and ±5%) and directionally start the ion generator to spray neutralizing airflow into the electrostatic abnormal area; all control commands are sent to the equipment controller via the industrial bus, while the yarn electrostatic status is monitored in real time to verify the effectiveness of the control; finally, a structured record containing batch number, timestamp, real-time rating (e.g., "excellent"), trend type (e.g., "fluctuation"), control command (e.g., "temperature and humidity +2℃") and execution status will be output.

[0104] This solution enhances the accuracy of quality control by providing differentiated regulation of yarn quality trends and outputting quality rating logs. It allows for targeted measures to address different quality trends, avoiding the pitfalls of a "one-size-fits-all" approach. This enables rapid containment of defect escalation, stabilization of quality fluctuations, and optimization of resource allocation, significantly improving the stability and consistency of yarn quality. Simultaneously, it strengthens quality traceability and process optimization. Complete logs provide a basis for tracing the root causes of quality problems and summarizing optimal process parameter combinations, facilitating continuous production improvement. Furthermore, it reduces overall production costs from multiple dimensions, including reducing defective products, lowering energy consumption and equipment wear, and minimizing large-scale rework.

[0105] Figure 3 A schematic diagram of a real-time quality rating system for yarn spinning, provided as an embodiment of this application, is shown below. Figure 3 As shown, a real-time quality rating system 300 for yarn textile manufacturing in this embodiment includes: a charge monitoring module 301, a two-dimensional rating module 302, a trend inference module 303, and a control execution module 304.

[0106] The charge monitoring module 301 is used to acquire a spatiotemporal charge dataset and generate a yarn electrostatic state dataset through a multimodal sensing fusion strategy. The dual-dimensional rating module 302 is used to identify yarn textile quality defects based on the yarn electrostatic state dataset and real-time acquired yarn morphology data, and to perform dual-dimensional dynamic rating according to the defect type and severity to generate a real-time yarn quality rating. The trend inference module 303 is used to analyze the changing trend of the real-time yarn quality rating and generate a subsequent quality change trend of the yarn. The control execution module 304 is used to execute corresponding process control instructions according to the subsequent quality change trend of the yarn and output a quality rating log.

[0107] Optionally, the charge monitoring module 301 is specifically used for: using a spatially deployed ring array of electrostatic sensors to collect spatiotemporal charge datasets on key points of the yarn in real time without contact; the spatiotemporal charge dataset includes real-time charge intensity values ​​and real-time charge distribution characteristics; using a high frame rate visual sensor to capture yarn morphology data and motion trajectory in real time; synchronously analyzing the spatiotemporal correlation between the real-time charge intensity values, the real-time charge distribution characteristics, and the yarn morphology data to identify abnormal electrostatic adsorption trajectories caused by quality defects; and generating the yarn electrostatic state dataset based on the spatiotemporal charge dataset, the yarn morphology data, and the abnormal electrostatic adsorption trajectories.

[0108] Optionally, when the dual-dimensional rating module 302 identifies yarn textile quality defects based on the yarn electrostatic state dataset and real-time collected yarn morphology data, and performs dual-dimensional dynamic rating according to the defect type and severity to generate an initial yarn quality rating, it is specifically used for: identifying yarn defect types based on the yarn electrostatic state dataset and real-time collected yarn morphology data, and determining the severity of the corresponding explicit defects for each yarn defect type; the yarn defect types include excessive hairiness, uneven yarn count, and fiber knots; normalizing several yarn defect types to a unified numerical range to determine the normalized yarn defect types; determining a dynamic rating score based on the normalized yarn defect types and the severity of the explicit defects; and generating a real-time yarn quality rating based on the dynamic rating score, wherein the real-time yarn quality rating is divided into three levels: excellent, medium, and poor.

[0109] Optionally, when the dual-dimensional rating module 302 identifies yarn defect types and determines the severity of the corresponding explicit defects based on the yarn electrostatic state dataset and the real-time acquired yarn morphology data, it is specifically used for: acquiring the yarn appearance image and the corresponding real-time charge intensity value in real time; and performing the following hair density gridding analysis steps based on the yarn appearance image: when the number of discrete fibers in several consecutive detection data exceeds a preset first density threshold and the average electrostatic intensity is greater than a preset multiple of the visual hair count, it is determined to be a local hair aggregation defect; when the number of discrete fibers in a unit detection area exceeding a preset proportion exceeds a preset second density threshold and the overall electrostatic distribution non-uniformity is greater than a preset diffuse hair proportion, it is determined to be a diffuse hair excess defect; and based on the combination pattern of the number of local aggregation areas and the diffuse distribution proportion, combined with the coupling coefficient between the electrostatic intensity increase and hair density, the spatial distribution state of hair is analyzed, and the explicitness of hair density is quantified.

[0110] Optionally, when the dual-dimensional rating module 302 identifies yarn defect types and determines the severity of explicit defects corresponding to the yarn defect types based on the yarn electrostatic state dataset and the real-time acquired yarn morphology data, it is specifically used for: acquiring a yarn axial image sequence in real time, and performing the following three-dimensional fluctuation analysis steps based on the yarn axial image sequence: marking abnormal diameter regions where the diameter difference between adjacent sampling points exceeds the diameter tolerance threshold as abrupt unevenness regions, simultaneously detecting the rate of change of the electrostatic adsorption trajectory direction angle of the abnormal diameter regions, and when the rate of change of the direction angle exceeds a preset turbulence threshold, the current abnormal diameter region is... The region is labeled as an electrostatic turbulence-type abrupt non-uniformity; the region corresponding to a continuous unidirectional increase / decrease in diameter exceeding a preset number of points is labeled as a gradual non-uniformity region. The slope of the electrostatic gradient distribution in the gradual region is obtained simultaneously. When the slope is in the same direction as the diameter change trend and exceeds the associated threshold, it is labeled as an electrostatic accumulation-type gradual non-uniformity. The proportions of the length of the abrupt non-uniformity region and the length of the gradual non-uniformity region in the unit detection area are determined. The degree of non-uniformity is quantified by combining the maximum diameter deviation value positively correlated with the degree of non-uniformity and the electrostatic turbulence intensity parameter, and the electrostatic gradient consistency parameter negatively correlated with the degree of non-uniformity.

[0111] Optionally, when the dual-dimensional rating module 302 identifies yarn defect types and determines the severity of visible defects corresponding to the yarn defect types based on the yarn electrostatic state dataset and the real-time acquired yarn morphology data, it is specifically used to: acquire real-time scanning imaging data of the yarn surface and simultaneously capture the corresponding electrostatic intensity extreme values, and perform the following impurity multi-scale feature extraction steps to clarify the impurity attributes: when there is a dark clump structure with irregular boundaries and a charge decay rate less than a preset decay rate threshold, the dark clump structure is determined to be a fiber agglomerate; when there is a reflective particle with geometric edges and an electrostatic intensity extreme value greater than a preset multiple of the particle size, the reflective particle is determined to be a hard impurity; when there is a radial fiber entanglement structure with an oscillating electrostatic intensity, the radial fiber entanglement structure is determined to be a yarn defect nodule; according to the size level, optical contrast, and boundary sharpness of the impurity multi-scale features, the size level, optical contrast, and boundary sharpness are weighted according to the impurity attributes to quantify the degree of fiber agglomeration visibility.

[0112] Optionally, when the dual-dimensional rating module 302 generates a real-time yarn quality rating based on the dynamic rating score, and the real-time yarn quality rating is divided into three levels: excellent, medium, and poor, it is specifically used for: mapping the identified yarn defect types to defect type identifiers based on a preset rule table, and mapping the assessed severity of the explicit defects to severity level identifiers; concatenating the defect type identifiers with the corresponding severity level identifiers, and using the concatenated result as a unit rating code; after all the unit rating codes are concatenated, performing a second string concatenation on the unit rating codes to generate a composite rating code; and mapping the composite rating code to an initial quality level of excellent / medium / poor according to a preset mapping rule. The mapping rules include: determining the total number of all defects as the defect density, and identifying the highest severity level identifier in the composite rating code; if the defect density is greater than a first density threshold or the highest severity level identifier reaches the first severity level threshold, it is judged as poor; if the defect density does not meet the aforementioned condition for judging as poor but is greater than a second density threshold or the highest severity level identifier reaches the second severity level threshold, it is judged as medium; if the defect density does not exceed the second density threshold and the highest severity level identifier does not reach the second severity level threshold, it is judged as excellent; wherein, the first density threshold is greater than the second density threshold, and the first severity level threshold is higher than the second severity level threshold.

[0113] Optionally, the trend deduction module 303 is specifically used for: extracting trajectory temporal dynamic features, including trajectory fluctuation frequency, directional consistency, and intensity variation amplitude, based on the electrostatic adsorption abnormal trajectory; identifying trajectory evolution patterns based on the dynamic features: when the trajectory fluctuation frequency is lower than a preset low-frequency threshold and the directional consistency is higher than a stable threshold, it is a stable and gradual trend; when the trajectory fluctuation frequency is higher than a preset high-frequency threshold and the intensity variation amplitude continues to increase, it is a deterioration and diffusion pattern; when the directional consistency is lower than a disorder threshold and the intensity variation amplitude decreases periodically, it is an intermittent oscillation pattern; mapping the trajectory evolution patterns to quality trends: the stable and gradual trend is mapped to a defect mitigation trend, the deterioration and diffusion pattern is mapped to a defect escalation trend, and the intermittent oscillation pattern is mapped to a quality fluctuation trend; and generating a subsequent quality change trend of the yarn based on the quality trends and the severity of the apparent defects.

[0114] Optionally, the control execution module 304 is specifically used to: perform differentiated control based on the type of subsequent quality change trend of the yarn: if it is determined to be a defect escalation trend, trigger linkage control to reduce spinning speed and increase environmental humidification; if it is determined to be a defect mitigation trend, maintain the current process and reduce detection density; if it is determined to be a quality fluctuation trend, dynamically adjust the environmental temperature and humidity and initiate static electricity neutralization in a targeted manner; associate the real-time quality rating of the yarn, the subsequent quality change trend of the yarn, and the control instructions with production line batches and time, and generate and output a quality rating log.

Claims

1. A real-time quality rating method for yarn spinning, characterized in that, include: Acquire a spatiotemporal charge dataset and generate a yarn electrostatic state dataset using a multimodal sensing fusion strategy; Based on the yarn electrostatic state dataset and combined with real-time collected yarn morphology data, yarn textile quality defects are identified, and a two-dimensional dynamic rating is performed according to the defect type and severity to generate a real-time yarn quality rating. Based on the real-time quality rating of the yarn, a trend analysis of the change of the real-time quality rating of the yarn is performed to generate the subsequent quality change trend of the yarn; The subsequent quality change trend of yarn refers to the predicted direction and extent of quality change of yarn in the subsequent spinning process based on the current electrostatic state, quality rating and weighted calibration results. Based on the subsequent quality change trend of the yarn, execute the corresponding process control instructions and output the quality rating log; The multimodal sensing fusion strategy includes: Using a spatially deployed ring array of electrostatic sensors, spatiotemporal charge datasets on key points of yarn are collected in real time in a non-contact manner. The spatiotemporal charge dataset includes real-time charge intensity values ​​and real-time charge distribution characteristics; Using a high frame rate vision sensor, yarn shape data and motion trajectory are captured in real time; The spatiotemporal correlation between the real-time charge intensity value, the real-time charge distribution characteristics, and the yarn morphology data is analyzed synchronously to identify abnormal electrostatic adsorption trajectories caused by quality defects. The yarn electrostatic state dataset is generated based on the spatiotemporal charge dataset, the yarn morphology data, and the electrostatic adsorption anomaly trajectory. The step of analyzing the trend of the real-time quality rating of the yarn based on the real-time quality rating and generating the subsequent quality change trend of the yarn includes: Based on the electrostatic adsorption anomaly trajectory, extract the trajectory temporal dynamic features including trajectory fluctuation frequency, directional consistency, and intensity variation amplitude; Based on the aforementioned dynamic features, trajectory evolution patterns are identified. When the trajectory fluctuation frequency is lower than a preset low-frequency threshold and the directional consistency is higher than a stable threshold, it is in a smooth and gradual mode. When the trajectory fluctuation frequency is higher than a preset high-frequency threshold and the intensity variation amplitude continues to increase, it is a deterioration and diffusion mode. When the directional consistency is below the disorder threshold and the intensity variation amplitude decreases periodically, it is an intermittent oscillation mode; The trajectory evolution pattern is mapped to a quality trend: the steady and gradual pattern is mapped to a defect mitigation trend, the deterioration and diffusion pattern is mapped to a defect escalation trend, and the intermittent oscillation pattern is mapped to a quality fluctuation trend. Based on the quality trend and the severity of obvious defects, a subsequent quality change trend of the yarn is generated.

2. The method according to claim 1, characterized in that, Based on the yarn electrostatic state dataset and combined with real-time collected yarn morphology data, yarn textile quality defects are identified, and a two-dimensional dynamic rating is performed according to the defect type and severity to generate an initial yarn quality rating, including: Based on the yarn electrostatic state dataset and the real-time collected yarn morphology data, the yarn defect type is identified, and the severity of the apparent defect corresponding to the yarn defect type is determined. The yarn defect types include excessive hairiness, uneven yarn, and fiber knots; The yarn defect types are normalized to a uniform numerical range to determine the normalized yarn defect types. A dynamic rating score is determined based on the normalized yarn defect type and the severity of the visible defect; Based on the dynamic rating score, a real-time yarn quality rating is generated, which is divided into three levels: excellent, medium, and poor.

3. The method according to claim 2, characterized in that, The process of identifying yarn defect types and determining the severity of visible defects corresponding to each yarn defect type, based on the yarn electrostatic state dataset and the real-time collected yarn morphology data, includes: Real-time acquisition of yarn appearance images and corresponding real-time charge intensity values; and based on the yarn appearance images, performing the following hair density gridding analysis steps: When the number of discrete fibers in several consecutive detection data exceeds the preset first density threshold and the average electrostatic intensity is greater than a preset multiple of the visual hair count, it is determined to be a local hair aggregation defect. When the number of discrete fibers in a unit detection area exceeds a preset second density threshold and the overall electrostatic distribution non-uniformity is greater than a preset diffuse fuzz ratio, it is judged as an excessive diffuse fuzz defect. Based on the combination pattern of the number of local clustered areas and the proportion of diffuse distribution, and combined with the coupling coefficient between electrostatic intensity increase and feather density, the spatial distribution state of feathers is analyzed, and the degree of feather density dominance is quantified.

4. The method according to claim 3, characterized in that, The process of identifying yarn defect types and determining the severity of visible defects corresponding to each yarn defect type, based on the yarn electrostatic state dataset and the real-time collected yarn morphology data, includes: Real-time acquisition of yarn axial image sequences; and based on the yarn axial image sequences, performing the following three-dimensional yarn evenness analysis steps: The diameter abnormal region where the diameter difference between adjacent sampling points exceeds the diameter tolerance threshold is marked as a sudden non-uniform region. The direction angle change rate of the electrostatic adsorption trajectory in the diameter abnormal region is detected simultaneously. When the direction angle change rate exceeds the preset turbulence threshold, the current diameter abnormal region is marked as an electrostatic turbulence sudden non-uniform region. The abnormal trend region corresponding to the unidirectional continuous increase / decrease in diameter exceeding the preset number of points is marked as a gradual non-uniform region. The slope of the electrostatic gradient distribution of the gradual region is obtained simultaneously. When the slope is in the same direction as the diameter change trend and exceeds the associated threshold, it is marked as an electrostatic accumulation gradual non-uniform region. The proportions of the lengths of abrupt and gradual uneven regions in the unit detection area are determined. The degree of unevenness is quantified by combining the maximum diameter deviation value and electrostatic disturbance intensity parameter which are positively correlated with the degree of unevenness, and the electrostatic gradient consistency parameter which is negatively correlated with the degree of unevenness.

5. The method according to claim 4, characterized in that, The process of identifying yarn defect types and determining the severity of visible defects corresponding to each yarn defect type, based on the yarn electrostatic state dataset and the real-time collected yarn morphology data, includes: Real-time acquisition of yarn surface scanning imaging data, and simultaneous capture of corresponding electrostatic intensity extreme values, followed by the impurity multi-scale feature extraction steps to clarify impurity attributes: When a dark, clump-like structure with irregular boundaries exists and its charge decay rate is less than a preset decay rate threshold, the dark, clump-like structure is determined to be a fibrous agglomerate. When there are reflective particles with geometric edges whose electrostatic intensity extreme value is greater than a preset multiple of the particle size, the reflective particles are determined to be hard impurities. When a radial fiber entanglement structure with oscillating electrostatic intensity exists, the radial fiber entanglement structure is determined to be a yarn defect nodule. Based on the size level, optical contrast, and boundary sharpness of the impurity multi-scale characteristics, the size level, optical contrast, and boundary sharpness are weighted according to the impurity attributes to quantify the degree of fiber knot impurity visibility.

6. The method according to claim 5, characterized in that, The process involves generating a real-time yarn quality rating based on the dynamic rating score. This real-time yarn quality rating is divided into three levels: Excellent, Medium, and Poor. Based on a preset rule table, the identified yarn defect types are mapped to defect type identifiers, and the assessed severity of the visible defects is mapped to severity level identifiers. Several defect type identifiers are concatenated with the corresponding severity level identifiers, and the concatenated result is used as the unit rating code. After all the unit rating codes are concatenated, a secondary string concatenation is performed on several of the unit rating codes to generate a composite rating code; According to the preset mapping rules, the composite rating code is mapped to an initial quality level of excellent / medium / poor; The mapping rules include: determining the total number of all defects as the defect density, and identifying the highest severity level identifier in the composite rating code; If the defect density is greater than the first density threshold or the highest severity level indicator reaches the first severity level threshold, it is judged as poor; If the defect density does not meet the aforementioned criteria for being classified as poor but is greater than the second density threshold, or if the highest severity level indicator reaches the second severity level threshold, then it is classified as medium. If the defect density does not exceed the second density threshold and the highest severity level indicator does not reach the second severity level threshold, it is judged as excellent; Wherein, the first density threshold is greater than the second density threshold, and the first severity level threshold is higher than the second severity level threshold.

7. The method according to claim 6, characterized in that, The process of executing corresponding process control instructions based on the subsequent quality change trend of the yarn and outputting a quality rating log includes: Differential control is implemented based on the subsequent quality change trend of the yarn: If the defect is determined to be escalating, then the linkage control is triggered to reduce the spinning speed and increase the ambient humidification. If the defect is determined to be mitigating, the current process is maintained and the detection density is reduced. If the aforementioned quality fluctuation trend is determined, the ambient temperature and humidity will be dynamically adjusted and electrostatic neutralization will be initiated in a targeted manner. The real-time quality rating of the yarn, the subsequent quality change trend of the yarn, and the control instructions are associated with the production line batch and time to generate and output a quality rating log.

8. A real-time quality rating system for yarn spinning, characterized in that, Applied to the method as described in any one of claims 1-7, comprising: The charge monitoring module is used to acquire spatiotemporal charge datasets and generate yarn electrostatic state datasets through a multimodal sensing fusion strategy. The dual-dimensional rating module is used to identify yarn textile quality defects based on the yarn electrostatic state dataset and real-time collected yarn morphology data, and to perform dual-dimensional dynamic rating according to the defect type and severity to generate a real-time yarn quality rating. The trend projection module is used to analyze the changing trend of the real-time quality rating of the yarn based on the real-time quality rating of the yarn, and generate the subsequent quality change trend of the yarn. The control execution module is used to execute corresponding process control instructions based on the subsequent quality change trend of the yarn, and output a quality rating log.

Images