Machine picking cotton whole process quality monitoring and spinning process intelligent optimization system

Abstract

The application discloses a machine-picked cotton whole-process quality monitoring and spinning process intelligent optimization system and relates to the technical fields of machine-picked cotton and spinning.The system is composed of a data acquisition unit, a data management unit, a model unit, an intelligent optimization unit and a closed-loop verification unit.The application takes the relationship between cotton quality and spinning quality as the main line, connects the links of planting and harvesting, ginning processing, raw cotton detection, scutcher, carding, combing, spinning and bobbin winding, constructs a unified quality index system of raw cotton indexes, process parameters, online quality and yarn quality, realizes the transformation from after-event inspection to process prevention and closed-loop optimization through online monitoring, offline detection, abnormal alarm, yarn quality prediction, foreign fiber collaborative control and process optimization prescription recommendation, promotes the cooperation of ginning mills and spinning mills under the same quality language, and improves the use stability of machine-picked cotton and the consistency of spinning quality.

Description

Technical Field

[0001] This invention relates to the field of machine-harvested cotton and spinning technology, specifically a system for monitoring the quality of machine-harvested cotton throughout the entire process and for intelligent optimization of spinning technology. Background Technology

[0002] Machine-harvested cotton has become the main form of cotton production. While machine harvesting has improved harvesting efficiency and ensured large-scale cotton supply, it has also made spinning mills face more complex raw material conditions: problems such as mixed seed cotton grades, residual film mixing, high impurity content, increased short fibers, fluctuations in the proportion of immature fibers, and damage from ginning processes will be amplified at each stage of the cleaning and carding process, carding, combing, spinning, and winding, and will eventually manifest as neps, fine and coarse areas, hairiness, strength fluctuations, yarn breaks, and fabric defects.

[0003] Traditional quality control typically separates cotton purchasing, public inspection, ginning, spinning, and finished product inspection. Ginners focus more on national standards, public inspection, and trading indicators, while spinning mills focus more on spinnability indicators provided by HVI, AFIS, impurity meter, foreign fiber meter, electric cleaning machine, and evenness meter. Different testing methods, sample sizes, and evaluation criteria lead to different quality judgments for the same batch of cotton at the upstream and downstream levels. In particular, indicators such as moisture regain, impurity content, ginning quality, foreign fibers, neps, and SCI are often more sensitive in actual use by spinning mills than at the ginning end. If a traceable relationship cannot be established between these indicators and yarn quality results, raw material claims, process adjustments, and product downgrading can only rely on experience-based judgments.

[0004] Therefore, there is an urgent need for a system that can connect the upstream and downstream of the cotton industry chain and realize full-process quality monitoring and intelligent optimization to solve problems such as large fluctuations in the quality of machine-harvested cotton, inconsistent quality judgment between upstream and downstream, and low efficiency in foreign fiber control. Summary of the Invention

[0005] This invention provides a system for monitoring the quality of machine-harvested cotton throughout the entire process and for intelligently optimizing the spinning process. The aim of this invention is to transform cotton quality issues from single-point detection into a continuous data chain. The system not only collects the test results of raw cotton and semi-finished products, but also integrates data from fiber separators, electronic yarn clearers, single-spindle monitoring, laboratory instruments, and environmental sensors, forming a closed loop from fiber to yarn, and from process settings to quality results. This invention combines the quality characteristics of machine-harvested cotton, the differences in perception between ginning and spinning quality, the concept of a big data-based quality system for the entire spinning process, and experimental data from the machine-harvested cotton carding process to propose the structure, functional modules, and application path of the system for monitoring the quality of machine-harvested cotton throughout the entire process and for intelligently optimizing the spinning process.

[0006] To achieve the above objectives, the present invention provides the following technical solution: The whole-process quality monitoring and intelligent optimization system for machine-harvested cotton spinning includes: The data acquisition unit is used to collect data from the entire process of machine-harvested cotton. The data governance unit, whose input end is connected to the output end of the data acquisition unit, performs batch unification and index standardization processing on the data of the entire process of machine-harvested cotton, and then constructs a unified quality index system. The model unit, which consists of a mechanistic model, a statistical model, and a knowledge rule model, is used to construct a multi-dimensional mapping relationship between raw cotton properties, process parameters, and yarn quality. The intelligent optimization unit has its input end connected to the output end of the data governance unit and the output end of the model unit. Based on the unified quality index system and the model unit, the intelligent optimization unit diagnoses the quality risks in the spinning process and then outputs a process optimization prescription. The closed-loop verification unit, whose input is connected to the output of the intelligent optimization unit, dynamically adjusts and verifies the process parameters based on the process optimization prescription and the full-process data of machine-harvested cotton.

[0007] As a preferred embodiment of the present invention, the machine-harvested cotton process data includes: Online data includes data from the foreign fiber machine, electronic yarn clearer shearing data, single spindle monitoring data, and workshop temperature and humidity data; Offline data includes HVI index data, AFIS index data, evenness meter data, tensile strength meter data, impurity content data, moisture regain data, and raw cotton quality data; Process data includes beater speed data, dust bar spacing data, drop rate data, roller spacing data, and twist data.

[0008] As a preferred embodiment of the present invention The specific process by which the data governance unit performs batch unification and index standardization processing on the entire process data of machine-harvested cotton includes: For indicators with large values ​​and high risk, the following formula is used for standardized calculation: , , In the formula, For the first Batch No. Items, and These are the historical mean and standard deviation of the indicator, respectively. A risk score between 0 and 1; For indicators with large values ​​and good quality, a reverse standardization method is used. The unified quality indicator system includes: The basic indicators include length, strength, micronaire value, maturity, short fiber content, impurity content, neps, foreign fiber content, and moisture regain. Process indicators include waste rate, cleaning efficiency, carding gap, combing waste rate, yarn breakage rate, winding shearing times, and single spindle efficiency. The results indicators include yarn evenness CV index, IPI index, hairiness index, strength index, twist index, yarn evenness spectrum index, and fabric evaluation index. Cost indicators include cotton waste, recycled yarn, foreign fiber machine spray rate, yarn splicing, yarn grade reduction, and claim risk.

[0009] As a preferred embodiment of the present invention, the intelligent optimization unit includes a raw cotton and ginning quality coordination module, a cleaning and combing process monitoring module, a carding and combing quality optimization module, a quality risk classification module, a spinning and winding online early warning module, a yarn quality prediction module, and a process prescription module.

[0010] As a preferred embodiment of the present invention, the raw cotton and ginning quality coordination module is used to construct a two-way feedback mechanism for ginning-spinning quality, and to drive the ginning process orientation change based on the two-way feedback mechanism for ginning-spinning quality.

[0011] As a preferred embodiment of the present invention, the carding and combing process monitoring module generates a recommended range for the carding and combing waste rate based on the unified quality index system, and corrects it using the detection results of cotton layer, cotton lint, and sliver. The carding and combing quality optimization module establishes a standardized process model based on the raw cotton grade, and automatically prompts for priority adjustment when the quality indicators of multiple processes deviate. The cleaning and combing process monitoring module and the carding and combing quality optimization module optimize the process based on the quality benefit-cost balance index. The calculation model of the quality benefit-cost balance index is as follows: , In the formula, , and The removal efficiency for cotton knots, impurities, and foreign fibers are respectively. This refers to the rate of waste material or cotton waste. For the increase in short fiber rate, For energy consumption or equipment load, , , , , , Weights are assigned based on product objectives.

[0012] In a preferred embodiment of the present invention, the quality risk classification module sets risk weights based on yarn count, application, customer requirements, and finishing method, and then implements risk classification based on the risk weights. The specific calculation model is as follows: , In the formula, As a comprehensive quality risk index, Weights are set according to product orientation, and , ; Weight Different settings are made for bleached fabrics, high-speed woven yarns, combed high-count yarns, or ordinary medium-coarse yarns, and the initial weight values ​​are iteratively corrected based on historical production data.

[0013] As a preferred embodiment of the present invention, the online early warning module for yarn spinning and winding performs coordinated control of dissimilar fibers based on the approximate relationship of fiber conservation and the total cost model for fiber control, wherein: Establish the approximate relationship of heterofiber conservation: , In the formula, This represents the risk of foreign fibers when raw materials enter the system. This refers to the amount of foreign fibers removed from seed cotton. To remove loose fibers and debris, This refers to the amount of material removed from the winding tube. For residual risk level, This is due to sampling and recognition errors; Establish a total cost model for controlling foreign fibers: , In the formula, Losses caused by accidental spraying, accidental cutting, or excessive cleaning. to These are the cost coefficients corresponding to each stage; The online early warning module for spinning and winding employs a combination of statistical process control and the knowledge rule model to provide anomaly warnings, specifically including: The EWMA model is used to smooth the short-period noise of the online quality index: ,in ; The CUSUM statistic is used to detect persistent shifts in the process mean: , , when or Exceeding the preset threshold An alarm is triggered at any time.

[0014] In a preferred embodiment of the present invention, the yarn quality prediction module is used to predict yarn appearance, weaving performance, and fabric surface risk, and the process prescription module is used to output the optimized process prescription, wherein: The yarn quality prediction module adopts a hierarchical modeling strategy, specifically including: When the sample size is less than a preset threshold, a multiple linear model that includes the interaction between raw cotton indicators and process parameters is used for prediction. The prediction formula is as follows: , In the formula, For the first Predicted yarn quality value Indicators of raw cotton With process parameters The interaction coefficient between them; When the sample size is greater than or equal to a preset threshold, a random forest model or a gradient boosting tree model is used to describe the nonlinear relationship. The prediction formula for the random forest model is as follows: , In the formula, For the first A tree of return, This represents the total number of regression trees; The process formulation module outputs the optimized process formulation based on a multi-objective optimization algorithm. The objective function of the multi-objective optimization is: , The constraints are: , ; In the formula, For the process parameters to be optimized, , , , , and These represent yarn evenness, common yarn defects, hairiness, number of times the yarn is cleaned and cut, quality cost, and strength, respectively. to The weights corresponding to each indicator; The system uses grid search, Bayesian optimization, or NSGA-II multi-objective evolutionary algorithm to solve the Pareto solution set, and then selects executable process optimization prescriptions from them.

[0015] This invention provides a system for monitoring the quality of machine-harvested cotton throughout the entire process and for intelligently optimizing the spinning process. It offers the following advantages:

[0016] 1. This invention establishes a unified quality index system for cotton quality and spinning quality: the system unifies public inspection, ginning end inspection, spinning mill re-inspection, online quality and yarn quality into the same batch data model, enabling traceability between impurities, moisture regain, neps, short fibers, foreign fibers, SCI, electro-cleaning cut and evenness quality. This innovation helps reduce disputes between ginning mills and spinning mills due to different testing standards, and can also promote upstream processing to improve according to downstream spinnability targets.

[0017] 2. This invention constructs a process parameter recommendation model for machine-harvested cotton. Traditional process optimization often relies on the experience of experienced workers, involves many adjustment objects, and has a long verification cycle. The system uses historical batch data, combing test data, and expert knowledge to form a recommendation logic for the waste rate of the combing and cleaning process, cylinder cover spacing, top comb tooth density, combing waste rate, spinning speed curve, and cleaning parameters. The recommendation result is not a fixed prescription, but dynamically changes with the grade of raw cotton, variety and use, yarn count, fabric requirements, and quality and cost targets.

[0018] 3. This invention forms a closed-loop mechanism of "alarm-diagnosis-prescription-verification-solidification". After the system alarms, it not only points out the abnormal indicators, but also provides possible causes, inspection paths and adjustment suggestions. After the process personnel implement the measures, the system continues to track changes in the efficiency of sliver, yarn and winding, solidifies effective measures into the knowledge base and marks ineffective measures as inapplicable. As cases accumulate, the system can gradually form the enterprise's own knowledge graph of machine-harvested cotton usage, reducing the loss of experience caused by personnel turnover.

[0019] 4. This invention promotes the forward shift of quality control to the planting and ginning stages. Problems such as foreign fibers from mulch film, insufficient maturity, and processing damage in machine-harvested cotton cannot be remedied solely at the spinning end due to high costs. The system can feed back residual foreign fibers, neps, and short fibers from the spinning end to the planting, harvesting, and ginning processes, guiding residual film treatment, colored mulch film identification, configuration of seed cotton foreign fiber machines and lint foreign fiber machines, processing moisture control, and lint cleaning strength optimization. This extends quality prevention from the spinning workshop to the upstream of the cotton industry chain. Attached Figure Description

[0020] Figure 1 This is a system framework diagram of the present invention; Figure 2 This is a framework diagram of the intelligent optimization unit in this invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0022] Please see Figure 1 The whole-process quality monitoring and intelligent optimization system for machine-harvested cotton spinning consists of a data acquisition unit, a data governance unit, a model unit, an intelligent optimization unit, and a closed-loop verification unit, as detailed below: A. Regarding the data acquisition unit: The data acquisition unit is used to collect data from the entire process of machine-harvested cotton. In the data acquisition unit, the system connects to HVI, AFIS, impurity meter, moisture regain detector, foreign fiber detector, evenness meter, tensile strength meter, electronic yarn clearer, single spindle monitoring, temperature and humidity sensor, and production management system, thereby obtaining data from the entire process of machine-harvested cotton. Obtaining data from the entire process of machine-harvested cotton is essential, as the quality of yarn is not affected by a single indicator of machine-harvested cotton quality, but is the result of the combined effects of fiber length, uniformity, short fiber rate, maturity, impurity content, neps, fiber fragments, foreign fibers, moisture regain, and processing damage. For example: When the short fiber content is high and the length uniformity is poor, the drafting mechanism has difficulty effectively controlling fibers of different lengths, the proportion of floating fibers increases, and the unevenness and thickness of the yarn increase. When the maturity is poor, the fiber strength is low, the rigidity is small, and it is easily damaged. Short fibers and cotton knots are more likely to be generated during the cleaning and combing process, and the uniformity of subsequent dyeing is affected. When the impurities and harmful defects are too high, the fiber movement will be disrupted during opening, cleaning, carding and stretching, resulting in thick knots, broken ends and fabric defects; The mapping relationship between machine-picked cotton quality indicators and spinning quality risk is shown in Table 1: Table 1: Mapping Relationship between Quality Indicators of Machine-Harvested Cotton and Spinning Quality Risk

[0023] SCI (Spinnability Index) can be used as a rapid evaluation index of the spinnability of raw cotton. Its empirical formula integrates factors such as strength, micronaire value, length, uniformity, reflectance, and yellowness, and can predict yarn strength and spinnability to a certain extent. For machine-harvested cotton, SCI should not be used in isolation, but should be analyzed in conjunction with indicators such as short fibers, neps, impurities, foreign fibers, maturity, and ginning moisture regain. If a batch of cotton has a high SCI but abnormal foreign fibers, short fibers, or neps, it may still result in efficiency losses in the cleaning and winding stages. If the SCI is average but the batch consistency is good and foreign fibers and harmful defects are controlled, stable yarn quality may be obtained through process optimization. In summary, the combined effect of the above-mentioned indicators can lead to inconsistent yarn quality. Therefore, this system needs to acquire data on the entire process of machine-harvested cotton through the data acquisition unit, and then have the data governance unit perform batch unification and indicator standardization processing to build a unified quality indicator system. The complete data for machine-harvested cotton should cover the above-mentioned indicators. More specifically, the complete data for machine-harvested cotton consists of online data, offline data, and process data, among which: Online data includes data from the fiber sorting machine, electronic yarn clearer shearing data, single spindle monitoring data, and workshop temperature and humidity data. Online data is continuous and timely, and can be used to detect sudden anomalies. Offline data includes HVI index data, AFIS index data, evenness meter data, tensile strength meter data, impurity content data, moisture regain data, and raw cotton quality data. Offline data is accurate and comparable and can be used to calibrate models. Process data includes beater speed data, dust bar spacing data, drop rate data, roller spacing data, and twist data. Process data is the basis for forming optimization suggestions. B. Regarding the data governance unit: From cotton bales to carding and combing lines, and then to spinning and winding, machine-harvested cotton often involves blending and batch overlap. Without a batch relationship table, it is difficult to accurately reflect the yarn quality to the raw cotton quality. The system should establish traceability relationships between cotton bales, cotton blending bales, carding and combing lines, sliver cans, roving, spinning spindle positions, winding spindle positions, and yarn batch numbers. For indicators with different units or different testing standards, conversion or calibration rules should be set to avoid misjudgments caused by simple comparisons. Therefore, the system needs to solve the problems of batch unification, indicator standardization, sample time alignment, and outlier handling through a data governance unit. Specifically: The input end of the data governance unit is connected to the output end of the data acquisition unit. The data governance unit performs batch unification and indicator standardization processing on the data of the entire machine-harvested cotton process, and then constructs a unified quality indicator system. The specific process of batch unification and indicator standardization processing of the data of the entire machine-harvested cotton process by the data governance unit includes: For indicators with large values ​​and high risk, the following formula is used for standardized calculation: , , In the formula, For the first Batch No. Items, and These are the historical mean and standard deviation of the indicator, respectively. A risk score between 0 and 1; For indicators with large values ​​and good quality, a reverse standardization method is used. This embodiment determines whether an indicator belongs to high-risk or high-quality indicators through indicator classification and standardization processing rules (see Table 2). Specifically, it adopts a three-step method of "process mechanism determination - historical data verification - customer limit review" to determine the indicators, and further classifies the indicators into three more specific categories: monotonically increasing risk type, monotonically decreasing risk type, and interval optimal type. The first step is to determine the direction based on the fiber processing mechanism: when an increase in a certain indicator will lead to an increase in neps, IPI, breakage, FD cutting times, fabric defects, or quality costs, it is classified as a monotonically increasing risk type; when an increase in a certain indicator will improve fiber spinnability, strength, uniformity, or production efficiency, it is classified as a monotonically decreasing risk type; when both excessively high and excessively low indicators will cause risks, it is classified as an interval optimal type. The second step involves verification using no fewer than 30 recent production batches of the same variety and yarn count, with the comprehensive yarn loss index as the benchmark. (The Spearman correlation coefficient was calculated using the standardized weighted average of stem CV, IPI, breakage rate, FD cut count, and customer rejection rate) as the dependent variable. Furthermore, if the direction is consistent in at least 80% of the product groups, it is considered a monotonically increasing risk type; When the correlation is U-shaped, or when there are upper and lower limits for customers or bidirectional loss of control in the mechanism, it is identified as the interval optimal type. When the sample is insufficient, the mechanism and customer standards are used to determine the initial direction first, and the automatic review is performed after a total of 30 batches. Third, customer quality specifications or equipment safety limits take precedence over statistical direction. Any batch that exceeds the hard limit will be directly subject to over-limit handling, and will no longer be judged solely based on standardized scores. Table 2: Indicator Classification and Standardization Rules

[0024] Where μ and σ are the mean and standard deviation of the same variety and yarn count during the stable production period; L and U are the lower and upper limits allowed by product standards or verified stable production; σ d For the standard deviation of historical deviations, the system's data dictionary stores "direction type, unit, applicable varieties, benchmark mean, standard deviation, upper and lower limits, data source and most recent revision date" for each indicator, thereby ensuring that the same indicator is consistent in direction during training, online monitoring and prescription recommendation. After the data from the entire process of machine-harvested cotton is processed as described above, a unified quality indicator system is constructed. This system consists of basic indicators, process indicators, result indicators, and cost indicators. Specifically: The basic indicators include length, strength, micronaire value, maturity, short fiber content, impurity content, neps, foreign fiber content, and moisture regain. Process indicators include waste rate, cleaning efficiency, carding gap, combing waste rate, yarn breakage rate, winding shearing times, and single spindle efficiency. The results indicators include yarn evenness CV index, IPI index, hairiness index, strength index, twist index, yarn evenness spectrum index, and fabric evaluation index. Cost indicators include cotton waste loss, recycled yarn, foreign fiber machine spray rate, yarn splicing, yarn grade reduction, and claim risk. C. Regarding model units: The model unit consists of a mechanistic model, a statistical model, and a knowledge rule model, used to construct a multi-dimensional mapping relationship between raw cotton properties, process parameters, and yarn quality, wherein: Mechanistic models reflect the influence of fiber length, short fibers, maturity, neps, impurities, and foreign fibers on drafting, carding, and yarn quality. Statistical models are used to establish the correlation between raw cotton indicators, process parameters, and yarn evenness CV, IPI, hairiness, strength, breakage, and number of cuts; The knowledge rule model comes from the experience of process engineers and historical cases, such as prioritizing the inspection of rubber rings, roller wrapping, roving twist and fly waste when there is a sudden increase in short and thick cuts. The combination of the three types of models enables the system to have both data learning capabilities and retain professional judgment in spinning. D. Regarding the intelligent optimization unit: Please see Figure 1 and Figure 2 The intelligent optimization unit's input is connected to the output of the data governance unit and the model unit. Based on a unified quality index system and model unit, the intelligent optimization unit diagnoses quality risks in the spinning process and then outputs a process optimization prescription. The intelligent optimization unit includes a raw cotton and ginning quality coordination module, a cleaning and combing process monitoring module, a carding and combing quality optimization module, a quality risk classification module, a spinning and winding online early warning module, a yarn quality prediction module, and a process prescription module. Specifically: D-1. Regarding the collaborative module for raw cotton and ginning quality: The raw cotton and ginning quality collaboration module is used to construct a two-way feedback mechanism for ginning and spinning quality. Based on this mechanism, the ginning process orientation is transformed. The quality discrepancies between the ginning and spinning ends are essentially due to different quality evaluation objectives. Ginning mills prioritize cotton appearance, grade, and transaction compliance; spinning mills prioritize stable spinnability, low breakage, low yarn defects, and predictable fabric finish. For example, regarding impurity content, ginning-end testing focuses more on cotton appearance or standard inspection, while the impurity meter used in spinning mills simulates the combing process, more closely reflecting the actual difficulty of fiber removal during the cleaning and combing process. Similarly, regarding ginning quality, there are significant differences in the accuracy of manually visually inspected defects compared to instrumental indicators such as AFIS neps and fiber fragments. The system integrates these... Different calibers are mapped to a unified quality label, preserving both ginning end indicators and spinning end feedback. This enables upstream and downstream companies to establish a common judgment on the same batch. Therefore, this module of the system records cotton variety, origin, harvesting method, seed cotton grade, moisture regain, ginning process, lint cleaning configuration, foreign fiber removal effect, public inspection indicators, and spinning mill re-inspection indicators. For items that are prone to disputes, such as impurity content, moisture regain, ginning quality, and foreign fibers, the system saves ginning end detection values, spinning end detection values, and difference ranges separately. Through long-term data accumulation, an evaluation of the impact of ginning parameters on neps, short fibers, and foreign fibers can be formed, promoting ginning mills to shift from a "white and clean" appearance orientation to a process orientation that also considers the usability of spinning. D-2. Regarding the monitoring module for the combing and cleaning process: The cleaning and carding process monitoring module generates a recommended range for the waste rate of the cleaning and carding process based on a unified quality index system, and corrects it through the detection results of cotton layer, cotton lint and sliver. Specifically, the cleaning and carding process is a key link in the transformation and release of quality risks of machine-picked cotton. The system should focus on monitoring the opening strength, waste rate, impurity removal efficiency, short fiber changes, nip changes, foreign fiber residue and dust removal status. For machine-picked cotton, the principles of gentle opening, early shedding and less breakage, and gradual opening should be adhered to. Excessive fiber beating should be avoided in order to improve the appearance of cleanliness. The system can provide suggested ranges for the waste rate of the combing and carding process and the speed of the main beaters based on the level of impurities, neps and short fibers in the raw cotton, and can be corrected by the test results of cotton layer, cotton lint and sliver. Foreign fiber residue, especially foreign fibers from mulch film, is a challenge for quality control of machine-harvested cotton. Mulch film is thin, soft, fragile, and light-colored. After harvesting, ginning, and opening, it often transforms from large film sheets into small fragments or fibrous foreign objects. Large areas of foreign fibers can be identified and removed by foreign fiber removal devices, but small mulch film foreign fibers may enter downstream with the fiber layer, causing an increase in the number of electro-cleaning cuts, a decrease in winding efficiency, and even unacceptable color defects on bleached or light-colored fabrics. Therefore, foreign fiber control cannot rely solely on end-of-line electro-cleaning and removal, but should be moved forward to field residual film management, seed cotton and lint foreign fiber removal, control of cleaning and carding waste, and evaluation of residual foreign fibers at the spinning end. D-3. Regarding the carding and combing quality optimization module: The carding and combing quality optimization module establishes a standardized process model based on the raw cotton grade. When quality indicators in multiple processes deviate, it automatically prompts the priority adjustment targets. Specifically, the carding process should focus on cylinder cover spacing, licker-in speed, carding element status, additional carding plates, and the configuration of impurity removal devices. The combing process should focus on lap weight, cylinder tooth density, top card tooth density, cotton waste rate, and combed sliver quality. The system can establish standardized process models for different raw cotton grades. When indicators such as neps, fiber scraps, short fibers, and foreign fibers deviate from the target, it automatically prompts the priority adjustment targets. For batches with high levels of residual foreign fibers from mulch film, the top card tooth density, cylinder needle density, and dust removal efficiency of the cleaning and carding unit should be used as linkage variables, rather than simply relying on the foreign fiber removal machine. In this embodiment, machine-harvested cotton was used as raw material. The waste rate during the cleaning and carding process was set at 2.5%, 3.5%, 4.5%, and 6.0%, respectively. The quality indicators of the cotton fibers were tested at each waste rate level. The experimental results are shown in Table 3. Table 3: Effect of the rate of waste material from carding and combing on the quality of cotton fibers in machine-picked cotton

[0025] Experimental results show that as the waste rate in the carding and combing process increases, the rates of neps, fiber-bearing seed dust, impurities, foreign matter, short fibers, immature fibers, and foreign fibers all decrease. When the waste rate in the carding and combing process increases from 2.5% to 6.0%, the number of neps decreases from 335 / g to 191 / g, fiber-bearing seed dust decreases from 112 / g to 45 / g, and foreign fibers decrease from 0.6 g / t to 0.2 g / t. This result indicates that moderately increasing the waste rate is effective in controlling harmful defects in machine-picked cotton, but it also means increased cleaning costs and cotton loss. Therefore, it is necessary to calculate the quality benefit and cost boundary. The monitoring module of the carding and combing process and the quality optimization module of carding and combing optimize the process based on the quality benefit-cost balance index. The calculation model of the quality benefit-cost balance index is as follows: , In the formula, , and The removal efficiency for cotton knots, impurities, and foreign fibers are respectively. This refers to the rate of waste material or cotton waste. For the increase in short fiber rate, For energy consumption or equipment load, , , , , , The specific details regarding the Quality Benefit-Cost Balance Index (hereinafter referred to as the G Index), which is weighted according to product objectives, are as follows: 1. The connection, calculation method, and judgment range between the G-index and the quality return equilibrium point: The financial break-even point is calculated as "fixed cost / (sales revenue - taxes - variable costs) × 100%". To ensure that the G-index in this specification can directly guide process adjustments, the original G-index is retained as the "technical benefit - damage cost" evaluation layer, and a quality benefit break-even point is added as the financial decision-making layer. For the i-th candidate process, the improvements in neps, impurities, foreign fibers, evenness, IPI, FD cutting count, strength, and winding efficiency are first converted into unit product quality benefits. in: For the benefit of quality dimension, For efficiency gains, For revenue based on grade level, For the benefit of complaints, losses from falling objects, energy consumption, machine downtime, needle cloth wear, and damage from excessive processing are converted into variable quality costs. ,in: For raw cotton cost, For energy costs, For process debugging costs, the amortization of testing, model operation, and routine maintenance are included as fixed quality costs. ,in: For the cost of the experiment, For process design costs, For equipment maintenance costs; The quality-benefit balance rate and the quality-benefit safety index used on site are respectively Among them, the smaller the BEPQ, the better, and the larger the SQ, the better. (equivalent to) To achieve the break-even point; when If the quality benefits cannot cover variable and fixed costs, it is directly judged as not meeting the revenue break-even condition, and the cost conversion uses the predicted price of raw cotton. and electricity price As a baseline for this implementation, the value of recycled waste cotton is deducted at 60% of the price of raw cotton. In formal application, the value is updated on a rolling basis according to the current financial price and the factory's internal ledgers. Table 4 shows the quality, benefit, and safety index grading thresholds and process handling rules, as detailed below: Table 4: Quality, Benefit and Safety Index Grading Thresholds and Process Handling Rules

[0026] When selecting the optimal process combination, first apply hard constraints: yarn strength must not be lower than the customer's lower limit, yarn evenness, IPI, and FD cut counts must not exceed the upper limit, waste rate must not exceed Rmax, ΔSFI must not exceed ΔSFImax, and equipment load and energy consumption must not exceed the safety upper limit. After passing the hard constraints, only [the following processes are retained]... Candidate combinations are selected and sorted from highest to lowest SQ. When the difference between the first and second place is less than 3 points, the one with lower BEPQ and lower drop rate is selected. When the difference is still less than 1 point, the one with lower variable quality cost and less equipment adjustment is selected. The recommended prescription must be verified by small batch comparison with the same batch of raw materials, adjacent machines or adjacent time windows. 2. Weighting coefficient calibration method and value standard: The weights were determined by a fusion of expert weights and data weights. Expert weights were obtained by pairwise comparisons using a scale of 1-9 by five engineers with over 10 years of experience in raw materials, carding, combing, spinning, and quality control, with a consensus ratio (CR) of less than 0.10. Data weights were derived from a quality cost regression model based on the most recent 180 consecutive production batches. The final weights were determined according to... The calculations are then normalized again to make the sum of the six weights equal to 1. These weights are used to convert each quality improvement into BQ (Balance of Quality), and losses, energy consumption, and damage into CV (Currency Value). Monetary parameters such as raw cotton price, electricity price, waste cotton recycling price, downgrade price difference, and customer losses are directly taken from the financial ledger. These parameters are recalibrated every 60 valid batches or when customer quality requirements change. The magnitude of a single correction should generally not exceed 20% of the original value. Table 5 uses 29.2tex carded unbleached yarn for weight calibration, and Table 6 shows the recommended initial weights for different product applications, as detailed below: Table 5: Weighting of 29.2tex Carded Yarn

[0027] Table 6: Recommended Initial Weights for Different Product Applications

[0028] When a customer specifies that "no visible foreign fibers should appear after bleaching," a foreign fiber correction factor should be applied. Set to 1.30; when high-count yarns are sensitive to neps and short fibers, and Set it to 1.20; for ordinary, low-grade yarn, the corresponding factor can be set to 0.90, and then adjusted accordingly. Renormalize to avoid changes in the total weight; 3. Calculation and verification of the quality benefit break-even point of the combing and cleaning process: Taking the production of 29.2tex ring-spun yarn from the same batch of Xinjiang machine-picked cotton as an example, the raw cotton has 420 neps / g, 125 impurities / g, 0.8g / t of foreign fibers, and a short fiber rate of 16.2%. Each combination operates continuously for 6 shifts. Cost accounting is based on... Calculate the net waste cost, where Pc is taken as 18,500 yuan / t and the waste cotton recovery rate γ is taken as 60%; electricity cost is calculated at 0.35 yuan / (kW·h); machine adjustment, downtime, and damage from excessive processing are calculated according to the factory's ledger. In this embodiment, the fixed quality cost is taken as... BQ is calculated based on reduced downgrade losses, improved winding efficiency, and reduced customer risk. CV consists of net waste cost, energy consumption, setup costs, and equipment damage. Table 7 shows the original calculation data for the candidate cleaning and carding processes, and Table 8 shows the calculation results of the break-even point for the candidate cleaning and carding processes, as detailed below: Table 7: Original Calculation Data for Candidate Processes for Combining Cleaning and Carding

[0029] Table 8: Calculation Results of the Break-Even Point of Candidate Processes for Combining Cleaning and Combing

[0030] The complete calculation for combination C is as follows: Variable quality cost The quality benefit (BQ) is 760 yuan / t, and the fixed quality cost (CF) is 60 yuan / t. Therefore... Scores above 65 enter the preferred selection zone. Although combination D has a higher clearing efficiency, The data is only within the feasible equilibrium range, and the waste rate and equipment damage costs have increased significantly. Therefore, 4.5% is the optimal balance point for the profitability of this batch of raw materials. Although 6.0% scores above 65 points, it is not the first choice. Table 9 compares the data between A, C, and D with the data in Tables 7 and 8 as a reference. Table 9: Results of the First Controlled Verification

[0031] 4. Calculation and verification of the break-even point for the yield of the top comb tooth density: Taking the production of 14.6tex compact spun bleached yarn from the same batch of raw materials as an example, the pre-combing cotton lap has 38 neps / g, 8 impurities / g, 0.6g / t of foreign fibers, and a short fiber rate of 11.2%, with target removal efficiencies set at 50%, 90%, and 75%, respectively; the combing waste baseline is 15.5%, with an upper limit of 19.0%; the energy consumption baseline is 107kW·h / t, with an upper limit of 120kW·h / t. The bleached product prioritizes the benefit weighting of foreign fiber residue and customer fabric risk, with fixed quality costs taking... BQ includes reductions in foreign fibers, neps, IPI, and downgrade losses; CV includes net combing waste, energy consumption, and top combing adjustment and maintenance costs. Table 10 presents the original calculation data for candidate combing processes, Table 11 presents the calculated break-even point results for candidate combing processes, and Table 12 provides a comparison with the data in Tables 10 and 11, as detailed below: Table 10: Original Calculation Data for Candidate Combing Processes

[0032] Table 11: Calculation Results of the Break-Even Point for Combing Candidate Processes

[0033] Table 12: Comparative Validation Results of Example 2 (Average of 6 Batches)

[0034] Combination G showed significant improvements in neps, short fiber rate, and foreign fiber content, while its cotton waste rate and energy consumption did not reach their upper limits. The highest value was 32 teeth / cm, therefore 32 teeth / cm was chosen as the recommended top comb density for this batch of raw materials. Although the yarn quality of the combination H continued to improve slightly, The decline in the safety index indicates that the marginal cost of increased cotton waste and energy consumption has exceeded the increased quality benefits, and it is not the first choice. The above two examples show that the combination of the financial benefit balance point and the original G index can not only ensure that the recommended schemes all reach 65 points or above, but also identify the inflection point of "quality continues to improve but the marginal cost has exceeded the benefits". Meanwhile, the setting of the carding and combing quality optimization module is essential. Since the removal effect of mulch-related foreign fibers also shows obvious process differences, as the grade of machine-picked cotton decreases, the amount of mulch-related foreign fibers in the foreign fiber remover, cotton opener, cotton cleaner, and dust collector all increases. Among them, the removal amount of cotton cleaner is the highest. This indicates that mulch-related foreign fibers are not entirely suitable for processing with a single image recognition device. The opening, impurity removal, and dust removal of the cleaning and carding unit have an important impact on their migration. The system should analyze the number of foreign fiber sprays, the number of FD cuts in the electro-cleaning unit, the amount of foreign fibers in the cotton cleaner, and the amount of foreign fibers in the dust collector on the same interface to determine whether the risk of foreign fibers should be moved forward or intercepted at the winding end. This embodiment demonstrates the improvement of machine-picked cotton quality through experiments on key adjustments to carding and combing. The direction of improvement in machine-picked cotton quality due to key adjustments to carding and combing is shown in Table 13. Table 13: Directions for Improving the Quality of Machine-Harvested Cotton Through Key Adjustments in Carding and Combing

[0035] Test data from the carding process show that reducing the cylinder flats spacing is most effective in improving the quality of carded slivers from machine-harvested cotton in Xinjiang. Under the listed adjustment schemes, neps can be reduced from 68 / g to 42 / g, fiber-bearing seeds from 19 / g to 11 / g, impurities from 28 / g to 15 / g, foreign fiber spray rate from 0.5% to 0.2%, and short fiber rate from 15.8% to 13.5%. However, reducing the spacing places higher demands on equipment condition, card cloth sharpness, and operation and maintenance. The system needs to simultaneously monitor the cotton web condition, cotton waste changes, and card cloth life to avoid new fiber damage caused by excessive carding. Test data from the combing process show that increasing the top comb tooth density is most effective in reducing harmful defects and foreign fibers, followed by increasing the cylinder tooth density. After increasing the top comb tooth density, the number of neps decreased from 28 / g to 16 / g, the amount of fiber-bearing seeds and impurities decreased to 0, the short fiber rate decreased from 9.8% to 7.9%, and the foreign fiber rate decreased from 0.4 g / t to 0.1 g / t. Therefore, for high-quality combed yarn or varieties with high requirements for fabric surface, the combing process is not only a short fiber removal process, but also a secondary control process for mulch-related foreign fibers and harmful defects. Therefore, this system realizes the linkage evaluation of combing waste cost and yarn quality benefits through the carding and combing quality optimization module. D-4. Regarding the quality risk classification module: The quality risk grading module sets risk weights based on yarn count, application, customer requirements, and finishing methods, and then achieves risk grading based on these risk weights. Specifically, different varieties have different sensitivities to the same quality fluctuations; the control focus should differ for carded medium-coarse yarn, combed compact yarn, high-speed weaving yarn, and bleaching yarn. The system can set risk weights according to yarn count, application, customer requirements, and finishing methods: bleached and light-colored fabrics have increased weights for foreign fibers, yellowish-whiteness, and maturity; high-speed weaving varieties have increased weights for strength, hairiness, and twist quality; and high-grade combed varieties have increased weights for short fibers, neps, and evenness. After risk grading, alarm thresholds and process prescriptions can better match the actual product requirements. The specific calculation model is as follows: , In the formula, As a comprehensive quality risk index, Weights are set according to product orientation, and , ; Weight Different settings are made according to the product uses of bleached fabrics, high-speed woven yarns, combed high-count yarns, or ordinary medium-coarse yarns, and the initial weight values ​​are iteratively corrected through historical production data. Comprehensive quality risk index The value ranges from 0 to 1. A higher value indicates a higher probability of yarn quality fluctuations caused by raw materials or processes. The grading threshold is determined using historical batch quantiles, customer out-of-limit results, and... OC analysis was used for joint calibration. The following examples used 180 consecutive production batches for backtracking. An "abnormal batch" was defined as any one of the following exceeding the product limit: evenness, IPI, strength, FD cut count, or customer fabric evaluation. Table 14 shows the comprehensive quality risk index classification thresholds and handling procedures, as follows: Table 14: Comprehensive Quality Risk Index Grading Thresholds and Handling Procedures

[0036] The linkage logic between R and the process formulation module is as follows: When Call the standard process library at the time; When the risk contribution is the highest, the indicator with the highest risk contribution is used as the prescription optimization target. For example, when the short fiber rate contributes the most, priority is given to adjusting the cotton blending, flexible opening, combing waste and roller spacing. At that time, the system calls Bayesian optimization or NSGA-II to generate no more than 3 sets of reversible prescriptions and requires single-machine verification; At this time, the prescription module only outputs suggestions for isolation, downgrading, or return, and does not automatically recommend enhanced processing, in order to avoid over-punishment and cost overruns; The linkage between R and CUSUM adopts a "dual-condition upgrade" rule. When a low-risk batch experiences a CUSUM alarm, the sensors, batch change, cleaning, and equipment status are checked first. When a medium-risk batch experiences an alarm, a cause list and compensation prescription are automatically generated. When a high-risk batch experiences an alarm, the automatic issuance of prescriptions is immediately suspended and manual confirmation is required. If any key indicator of a severely risky batch alarms, the batch is stopped from entering the target product. The risk level is recalculated once per cotton blending cycle, and online process indicators are updated every 5 to 30 minutes. D-5. Regarding the online early warning module for spinning and winding: The monitoring of individual spinning spindles, electronic yarn clearing and cutting cycles, bobbin forming curves, and environmental data can reflect the location and time of quality anomalies. The system should establish alarm rules according to variety, machine, spindle position, batch, and time, and implement tiered management for short / thick, long / thick, thin, neps, hairiness, yarn breakage, twisting anomalies, and inefficient spindles. Compared to traditional manual inspection, online early warning can pinpoint problems to the machine, spindle position, and parameter variation range, and, combined with a knowledge base, provide possible causes such as damaged rubber rings, roller entanglement, nozzle snagging, temperature and humidity fluctuations, poor bobbin forming, or excessive spinning tension. Specifically: The online early warning module for spinning and winding performs coordinated control of dissimilar fibers based on the approximate relationship of fiber conservation and the total cost model for fiber control. Specifically, it includes: Step a: Establish the approximate conservation relationship for heterofibers: , In the formula, This represents the risk of foreign fibers when raw materials enter the system. This refers to the amount of foreign fibers removed from seed cotton. To remove loose fibers and debris, This refers to the amount of material removed from the winding tube. For residual risk level, This is due to sampling and recognition errors; Step b: Establish a total cost model for foreign fiber control: , In the formula, Losses caused by accidental spraying, accidental cutting, or excessive cleaning. to The cost coefficients for each stage are defined; the system achieves coordinated control of foreign fibers between upstream and downstream processes by minimizing the total cost model for foreign fiber control. The online early warning module for spinning and winding employs a combination of statistical process control and knowledge rule models for anomaly warning, specifically including: The EWMA model is used to smooth the short-period noise of the online quality index: ,in , This is the "single online quality monitoring value" at the t-th sampling time, not a direct mixture of multiple indicators with different dimensions. The j-th indicator can be written as... The system establishes an EWMA sequence for each indicator; when cross-indicator integration is required, the direction is first unified and standardized, and then a comprehensive online risk value is constructed. In the default implementation, where The original measurement value output by the online device in the t-th time window. and To stabilize the mean and standard deviation during the production period, Directional coefficient: the larger the value, the higher the risk. When the value is higher, the quality is better. After standardization This indicates that the risk has shifted in an unfavorable direction. If the unit and range of the same indicator remain unchanged, the original value can also be used directly. As However, alarm thresholds must use the same unit, as shown in Table 15. The specific meaning and sampling method are as follows: Table 15: Specific meaning and sampling method

[0037] EWMA initial value Take the average value during the stable production period; use standardized q. t hour For 5-minute high-frequency data, the default is... The default setting is for 30-minute or shift data. The smaller the λ, the stronger the noise reduction capability but the slower the response. If a batch change, downtime exceeding 30 minutes or instrument calibration occurs, the system will shut down updates during that time period and resume after stabilizing 10 sampling points. The CUSUM statistic is used to detect persistent shifts in the process mean: , , when or Exceeding the preset threshold An alarm is triggered in a timely manner, and the alarm information is matched with the cause, inspection path, and handling measures in the knowledge base to form a closed-loop control of "alarm-diagnosis-prescription-verification-solidification". Specifically: The CUSUM alarm threshold h is preset in three steps: "stable baseline - minimum detectable offset - false alarm rate backtesting". First, select no fewer than 30 stable production batches or no fewer than 200 continuous valid sampling points, excluding periods of batch change, machine downtime, cleaning, sensor calibration, and known equipment failure, and calculate the baseline mean. and standard deviation ; Next, the process engineers determine the minimum meaningful offset δ that needs to be detected, which is generally taken as... Reference value ; Finally, set h to Among them, general indicators High-risk batches, bleaching products, or key customer indicators If standardization is used ,but All of these are dimensionless values; The threshold is not the customer's acceptable limit, but rather the cumulative alarm limit for continuous deviation. The system must also retain the customer's hard limit: an alarm should be triggered immediately if a single point exceeds the hard limit; if the single point limit is not exceeded but the cumulative CUSUM exceeds h, the average value of the judgment process will continuously deviate. Before the threshold goes live, backtesting should be performed using the most recent complete cotton distribution cycle, requiring that false alarms during the stable period not exceed 1 per 200 sampling points, and... The detection rate of the simulated offset within 8 sampling points is not less than 90%; if this condition is not met, then... Adjustments can be made within a certain range, specifically: Table 16: CUSUM Threshold Preset

[0038] In the table above, FD cuts, head breaks, and CVm are represented by the upper CUSUM: Winding efficiency is achieved using the lower CUSUM: ,when or An alarm is triggered when the cumulative amount reaches h. Furthermore, if two out of three consecutive sampling points continue to change in an unfavorable direction, the system will issue a warning; when the alarm reaches h, it will be upgraded to a formal alarm. After the handling is completed, the cumulative amount can only be reset when 10 consecutive sampling points return to the baseline range. D-6. Regarding the yarn quality prediction module: The yarn quality prediction module is used to predict yarn appearance, weaving performance, and fabric surface risks. The process prescription module is used to output the optimized process prescription. Specifically, this module combines data from the evenness meter, tensile strength meter, electronic yarn clearer shearing data, single-spindle monitoring data, and raw cotton quality data to predict yarn appearance, weaving performance, and fabric surface risks. The system output should not stop at a "pass / fail" judgment, but should provide an executable process optimization prescription, including suggestions on cotton blending adjustments, carding and combing waste rate, carding gap, combing waste rate, spinning speed curve, and winding and clearing parameters. Each prescription should include applicable raw material boundaries, expected improvement indicators, possible side effects, and verification methods to facilitate small-batch verification by process engineers. The yarn quality prediction module adopts a hierarchical modeling strategy, which specifically includes: When the sample size is less than a preset threshold, a multiple linear model that includes the interaction between raw cotton indicators and process parameters is used for prediction. The prediction formula is as follows: , In the formula, For the first Predicted yarn quality value Indicators of raw cotton With process parameters The interaction coefficient between them; When the sample size is greater than or equal to a preset threshold, a random forest model or a gradient boosting tree model is used to describe the nonlinear relationship. The prediction formula for the random forest model is as follows: , In the formula, For the first A tree of return, This represents the total number of regression trees; D-7. Regarding the process formulation module: The process formulation module outputs optimized process formulations based on a multi-objective optimization algorithm. The objective function of the multi-objective optimization is: , The constraints are: , ; In the formula, For the process parameters to be optimized, , , , , and These represent yarn evenness, common yarn defects, hairiness, number of times the yarn is cleaned and cut, quality cost, and strength, respectively. to The weights corresponding to each indicator; The system uses grid search, Bayesian optimization, or NSGA-II multi-objective evolutionary algorithm to solve the Pareto solution set and selects executable process optimization prescriptions from it; E. Regarding the closed-loop verification unit: The input end of the closed-loop verification unit is connected to the output end of the intelligent optimization unit. Based on the process optimization prescription and combined with the whole process data of machine-harvested cotton, the process parameters are dynamically adjusted and verified. When the closed-loop verification unit dynamically adjusts and verifies the process parameters, it adopts a small-batch, rollback, and pairwise comparison method. For the same batch of raw materials, select similar machines or adjacent time windows as control groups to compare the neps in sliver, short fiber rate in combed sliver, yarn breakage, winding cuts and yarn strength before and after adjustment; under the premise of recording temperature and humidity, equipment status, shift and cleaning cycle covariates, when the quality indicators improve and the cost does not exceed the constraints, write the corresponding process optimization prescription into the enterprise-level process library. This invention is based on the concept of "data acquisition, quality modeling, anomaly diagnosis, process recommendation, and closed-loop verification". It then constructs a whole-process quality monitoring and intelligent optimization system for machine-harvested cotton spinning process by "data acquisition unit", "data governance unit", "model unit", "intelligent optimization unit" and "closed-loop verification unit". The application of this system is not limited to the spinning workshop, but extends the quality boundary forward to the planting, harvesting and ginning links, and backward to the winding, yarn forming and fabric quality feedback. This invention focuses on the relationship between cotton quality and spinning quality, integrating planting and harvesting, ginning, raw cotton testing, carding and combing, carding, combing, spinning, and winding to construct a unified data model for raw cotton indicators, process parameters, online quality, and yarn quality. Through online monitoring, offline detection, anomaly alarms, yarn quality prediction, foreign fiber synergistic control, and process prescription recommendation, it transforms quality issues from post-event inspection to process prevention and closed-loop optimization. Combined with data from machine-harvested cotton carding experiments and spinning quality management cases, it shows that indicators such as carding and combing waste rate, carding cylinder cover spacing, combing top comb tooth density, and winding electric cleaning cuts can serve as key variables for closed-loop quality control. It is believed that this system can promote collaboration between ginning mills and spinning mills under the same quality language, improving the stability of machine-harvested cotton use and the consistency of spinning quality. In summary, this system has broad industrial applicability: First, establish a unified quality index system for cotton quality and spinning quality: the system unifies public inspection, ginning end inspection, spinning mill re-inspection, online quality and yarn quality into the same batch data model, so that impurities, moisture regain, neps, short fibers, foreign fibers, SCI, electric cleaning cut and evenness quality can be traced back to each other. This innovation helps to reduce disputes between ginning mills and spinning mills due to different testing standards, and can also promote upstream processing to improve in accordance with downstream spinnability targets; Second, we construct a process parameter recommendation model for machine-harvested cotton. Traditional process optimization often relies on the experience of experienced workers, involves many adjustment objects, and has a long verification cycle. The system uses historical batch data, combing test data, and expert knowledge to form a recommendation logic for the waste rate of combing and cleaning, cylinder cover spacing, top comb tooth density, combing waste rate, spinning speed curve, and cleaning parameters. The recommendation result is not a fixed prescription, but changes dynamically with the grade of raw cotton, variety and use, yarn count, fabric requirements, and quality and cost targets. Third, a closed-loop mechanism of "alarm-diagnosis-prescription-verification-solidification" is formed. After the system alarms, it not only points out the abnormal indicators, but also provides possible causes, inspection paths and adjustment suggestions. After the process personnel implement the measures, the system continues to track changes in the efficiency of sliver, yarn, and winding, solidifies effective measures into the knowledge base, and marks ineffective measures as inapplicable. As cases accumulate, the system can gradually form the company's own knowledge graph of machine-harvested cotton usage, reducing the loss of experience caused by personnel turnover. Fourth, it promotes the forward shift of quality control to the planting and ginning stages. Problems such as foreign fibers from mulch film, insufficient maturity, and processing damage in machine-harvested cotton cannot be remedied solely at the spinning end due to high costs. The system can provide feedback on residual foreign fibers, neps, and short fibers at the spinning end to the planting, harvesting, and ginning processes, guiding the treatment of residual film, identification of colored mulch film, configuration of seed cotton foreign fiber machines and lint foreign fiber machines, control of processing moisture regain, and optimization of lint cleaning strength, thus extending quality prevention from the spinning workshop to the upstream of the cotton industry chain. Fifth, establish an evaluation system oriented towards application effectiveness. The effectiveness of system construction should not only be judged by the number of devices connected or the number of reports generated, but also by whether quality losses have been reduced, process response has been accelerated, and batch stability has been improved. It is recommended to use "three rates, two differences, and one cost" as the core evaluation criteria: anomaly detection rate, first-time handling efficiency, rate of reduction in inefficient spindles, batch-to-batch quality difference and spindle-to-spindle quality difference, and unit quality cost of qualified yarn. This evaluation system can link digital construction with actual business results and prevent the system from remaining at the display level. The implementation can be carried out in three steps: The first step is to build a data foundation, complete the data access of laboratory instruments, online equipment and production systems, and establish batch traceability and quality indicator dictionaries; the second step is to build typical scenario applications, prioritizing the implementation of raw cotton quality profiling, cleaning and combing efficiency, carding and combing quality trends, inefficient spindle identification and foreign fiber collaborative control. The third step is to build an intelligent optimization module, conduct small-batch trial spinning verification for key varieties, and gradually form an enterprise-level process prescription library. Each step should set clear evaluation indicators, such as yarn evenness CV, IPI, hairiness, strength, breakage, number of cuts, production efficiency, cotton waste rate and cotton cost. The core of quality control for machine-harvested cotton is not to simply attribute problems to raw cotton, nor to infinitely strengthen opening, combing, and yarn cleaning at the spinning end. Instead, it is to establish a quality causal chain from planting and harvesting to finished yarn and fabric. The machine-harvested cotton full-process quality monitoring and intelligent optimization system for spinning process proposed in this invention is based on the mapping between raw cotton quality and finished yarn quality, uses online detection and offline laboratory data fusion as a means, and aims at process optimization prescription recommendation and closed-loop verification. It can shift quality management from experience-based judgment to data-driven. Variables such as the waste rate in the combing and cleaning process, the carding cylinder cover spacing, the combing top comb tooth density, the number of foreign fiber sprays, the number of electro-cleaning cuts, and the spinning speed curve can serve as key levers for system optimization. This invention, through continuous monitoring and coordinated adjustment of these variables, can reduce the risks of neps, short fibers, harmful defects, and foreign fibers while controlling costs, thereby improving the stability of machine-picked cotton use, winding efficiency, and yarn quality consistency. In the future, we should further strengthen the sharing of upstream ginning processing data and spinning quality data, and improve the model parameters for different varieties, yarn counts, and fabric applications, so that the quality control of machine-picked cotton can move from single-plant optimization to supply chain collaborative optimization. The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A system for monitoring the quality of machine-harvested cotton throughout the entire process and for intelligently optimizing the spinning process, characterized in that: include: The data acquisition unit is used to collect data from the entire process of machine-harvested cotton. The data governance unit, whose input end is connected to the output end of the data acquisition unit, performs batch unification and index standardization processing on the data of the entire process of machine-harvested cotton, and then constructs a unified quality index system. The model unit, which consists of a mechanistic model, a statistical model, and a knowledge rule model, is used to construct a multi-dimensional mapping relationship between raw cotton properties, process parameters, and yarn quality. The intelligent optimization unit has its input end connected to the output end of the data governance unit and the output end of the model unit. Based on the unified quality index system and the model unit, the intelligent optimization unit diagnoses the quality risks in the spinning process and then outputs a process optimization prescription. A closed-loop verification unit, whose input is connected to the output of the intelligent optimization unit, dynamically adjusts and verifies the process parameters based on the process optimization prescription and the full process data of machine-harvested cotton. The intelligent optimization unit includes a raw cotton and ginning quality coordination module, a combing and cleaning process monitoring module, a carding and combing quality optimization module, a quality risk classification module, a spinning and winding online early warning module, a yarn quality prediction module, and a process prescription module. The carding and combing process monitoring module generates a recommended range for the waste rate based on the unified quality index system, and corrects it using the cotton layer, cotton lint, and sliver test results. The carding and combing quality optimization module establishes a standardized process model based on the raw cotton grade, and automatically prompts for priority adjustment when multiple process quality indicators deviate. The cleaning and combing process monitoring module and the carding and combing quality optimization module optimize the process based on the quality benefit-cost balance index. The calculation model of the quality benefit-cost balance index is as follows: ， In the formula, , and The removal efficiency for cotton knots, impurities, and foreign fibers are respectively. This refers to the rate of waste material or cotton waste. For the increase in short fiber rate, For energy consumption or equipment load, , , , , , Weights are assigned based on product objectives.

2. The intelligent optimization system for whole-process quality monitoring and spinning technology of machine-harvested cotton according to claim 1, characterized in that, The complete process data for machine-harvested cotton includes: Online data includes data from the foreign fiber machine, electronic yarn clearer shearing data, single spindle monitoring data, and workshop temperature and humidity data; Offline data includes HVI index data, AFIS index data, evenness meter data, tensile strength meter data, impurity content data, moisture regain data, and raw cotton quality data; Process data includes beater speed data, dust bar spacing data, drop rate data, roller spacing data, and twist data.

3. The whole-process quality monitoring and intelligent optimization system for machine-harvested cotton spinning as described in claim 1, characterized in that, The specific process by which the data governance unit performs batch unification and index standardization processing on the entire process data of machine-harvested cotton includes: For indicators with large values ​​and high risk, the following formula is used for standardized calculation: ， ， In the formula, For the first Batch No. Items, and These are the historical mean and standard deviation of the indicator, respectively. A risk score between 0 and 1; For indicators with large values ​​and good quality, a reverse standardization method is used. The unified quality indicator system includes: The basic indicators include length, strength, micronaire value, maturity, short fiber content, impurity content, neps, foreign fiber content, and moisture regain. Process indicators include waste rate, cleaning efficiency, carding gap, combing waste rate, yarn breakage rate, winding shearing times, and single spindle efficiency. The results indicators include yarn evenness CV index, IPI index, hairiness index, strength index, twist index, yarn evenness spectrum index, and fabric evaluation index. Cost indicators include cotton waste, recycled yarn, foreign fiber machine spray rate, yarn splicing, yarn grade reduction, and claim risk.

4. The whole-process quality monitoring and intelligent optimization system for machine-harvested cotton and spinning process according to claim 1, characterized in that, The raw cotton and ginning quality coordination module is used to construct a two-way feedback mechanism for ginning and spinning quality, and to drive the transformation of the ginning process based on the two-way feedback mechanism for ginning and spinning quality.

5. The whole-process quality monitoring and intelligent optimization system for machine-harvested cotton and spinning process according to claim 1, characterized in that, The quality risk grading module sets risk weights based on yarn count, intended use, customer requirements, and finishing methods, and then achieves risk grading based on these risk weights. The specific calculation model is as follows: ， In the formula, As a comprehensive quality risk index, Weights are set according to product orientation, and , ; Weight Different settings are made for bleached fabrics, high-speed woven yarns, combed high-count yarns, or ordinary medium-coarse yarns, and the initial weight values ​​are iteratively corrected based on historical production data.

6. The whole-process quality monitoring and intelligent optimization system for machine-harvested cotton and spinning process according to claim 1, characterized in that, The online early warning module for spinning and winding is based on the approximate relationship of foreign fiber conservation and the total cost model for foreign fiber control to perform collaborative control of foreign fibers, wherein: Establish the approximate relationship of heterofiber conservation: ， In the formula, This represents the risk of foreign fibers when raw materials enter the system. This refers to the amount of foreign fibers removed from seed cotton. To remove loose fibers and debris, This refers to the amount of material removed from the winding tube. For residual risk level, This is due to sampling and recognition errors; Establish a total cost model for controlling foreign fibers: ， In the formula, Losses caused by accidental spraying, accidental cutting, or excessive cleaning. to These are the cost coefficients corresponding to each stage; The online early warning module for spinning and winding employs a combination of statistical process control and the knowledge rule model to provide anomaly warnings, specifically including: The EWMA model is used to smooth the short-period noise of the online quality index: ,in: , This represents the single online quality monitoring value at the t-th sampling time. The CUSUM statistic is used to detect persistent shifts in the process mean: ， ， when or Exceeding the preset threshold An alarm is triggered at any time.

7. The whole-process quality monitoring and intelligent optimization system for machine-harvested cotton and spinning process according to claim 1, characterized in that, The yarn quality prediction module is used to predict yarn appearance, weaving performance, and fabric surface risk; the process prescription module is used to output the optimized process prescription, wherein: The yarn quality prediction module adopts a hierarchical modeling strategy, specifically including: When the sample size is less than a preset threshold, a multiple linear model that includes the interaction between raw cotton indicators and process parameters is used for prediction. The prediction formula is as follows: ， In the formula, For the first Predicted yarn quality value Indicators of raw cotton With process parameters The interaction coefficient between them; When the sample size is greater than or equal to a preset threshold, a random forest model or a gradient boosting tree model is used to describe the nonlinear relationship. The prediction formula for the random forest model is as follows: ， In the formula, For the first A tree of return, This represents the total number of regression trees; The process formulation module outputs the optimized process formulation based on a multi-objective optimization algorithm. The objective function of the multi-objective optimization is: ， The constraints are: , ; In the formula, For the process parameters to be optimized, , , , , and These represent yarn evenness, common yarn defects, hairiness, number of times the yarn is cleaned and cut, quality cost, and strength, respectively. to The weights corresponding to each indicator; The system uses grid search, Bayesian optimization, or NSGA-II multi-objective evolutionary algorithm to solve the Pareto solution set, and then selects executable process optimization prescriptions from them.

Images