Cotton spinning fiber quality online monitoring method and system based on intelligent sensor

Abstract

This invention provides a method and system for online monitoring of cotton fiber quality based on intelligent sensors, belonging to the field of textile production monitoring technology. It performs non-contact, multi-parameter synchronous detection of cotton fibers, converting the acquired local quality parameter vectors into width statistical feature quantities with macroscopic spatial semantics. This achieves matching between the spatial scale of the detection results and the scale of process parameter control, improving the accuracy of process adjustments. A database of quality parameter evolution, including quality transfer relationships between processes and thermodynamic energy balance, is established, and the process quality transfer function of each process is fitted. The sensitivity matrix of the final product quality parameters to the process parameters of each process is calculated using a perturbation method. Then, the globally optimal process parameter adjustment amount that minimizes the weighted quality deviation is solved, enabling spatial localization of quality anomalies and joint process optimization across processes. This achieves global coordinated optimization of process parameters across multiple processes, improving the systematicness and effectiveness of the entire process quality control.

Description

Technical Field

[0001] This invention relates to the field of textile production monitoring technology, specifically to a method and system for online monitoring of cotton fiber quality based on intelligent sensors. Background Technology

[0002] In cotton spinning production, online monitoring of fiber quality parameters is a core element in ensuring the quality of the final yarn. The cotton spinning process mainly includes opening and cleaning, carding, drawing, roving, and spinning. In each process, the fiber undergoes mechanical actions such as opening, impurity removal, combing, drafting, and twisting, and its quality state continuously evolves with the processing. The quality parameters that need to be monitored online mainly include: fiber maturity (reflecting the degree of cell wall thickening, directly related to fiber strength, elasticity, and cohesion), moisture content (affecting fiber spinnability and static electricity accumulation), short fiber rate (the proportion of fibers shorter than 16mm, affecting yarn strength and evenness), neps content (affecting yarn appearance and breakage rate), impurity content (affecting the cleanliness of the final product), and fiber fineness. In the prior art, CN121275676A discloses a method for simultaneous detection of multiple performance indicators of cotton textile yarns, which performs layered performance analysis by acquiring multi-layer structural parameters of the yarn body; CN121740880A discloses an online monitoring system for textile fabric production based on intelligent manufacturing, which calculates anomaly scores by constructing a multi-dimensional feature matrix and combining it with a neural network. Other related technical solutions include using near-infrared spectroscopy to analyze fiber composition, using microwave technology to detect moisture content, and using machine vision to detect cotton knots and impurities. However, the prior art has the following technical problems in its use: Problem 1: There is a mismatch in scale between the spatial scale of the detection parameters and the scale of process parameter control. Existing sensors are limited to a small local area within their field of view, reflecting only the quality status of a single point on the fiber web. However, process control targets macroscopic overall machine parameters (such as draft ratio and pressure), covering the entire fiber web. When a sensor detects a quality anomaly at a local location, current technology cannot determine whether the anomaly is localized or global, nor can it determine whether process parameters should be fine-tuned locally or adjusted globally. This spatial scale mismatch makes it difficult for detection results to directly and effectively guide process parameter adjustments. Problem 2: The quality monitoring systems of each process are independent of each other, lacking a joint optimization mechanism between quality and process. In existing technologies, each process outputs its own quality evaluation results based on its own quality monitoring system, and adjusts its process parameters accordingly. This results in the inability to correlate the quality status data of the same batch of fibers across different processes. When quality problems occur in the final product, it is impossible to accurately trace which process caused the problem. When a quality anomaly is detected in a subsequent process, it is impossible to determine whether the anomaly was input from a previous process or newly generated in the current process. The process parameter adjustments of each process are independent and lack global coordination. This leads to a series of interconnected effects from process parameter adjustments that cannot be quickly detected and located, and it cannot support joint process optimization across processes. Summary of the Invention

[0003] To achieve the above objectives, the present invention provides the following technical solution: a method and system for online monitoring of cotton fiber quality based on intelligent sensors, the method comprising: The capacitance and loss tangent of the fiber under multi-band alternating voltage are collected, and the fiber temperature, fiber layer thickness and conveying speed are collected simultaneously to form a detection record. The local quality parameter vector is obtained by solving the detection record. Obtain local quality parameter vectors for multiple channels in the fiber width direction, construct a width distribution vector and generate anomaly channel identification vectors, calculate width statistical features based on the width distribution vector, and output an outlet quality characterization data packet; Based on the exit quality characterization data package of the previous process, calculate the inlet quality parameter vector of the next process, obtain the process thermodynamic state and process operating parameters of each process, establish a quality parameter evolution database, fit the process quality transfer function, and calculate the amount of quality deviation generated in each process. When the final product quality parameters deviate from the preset target parameters, reverse trace of the process is performed to determine whether there is a process quality abnormality. If so, an abnormality location result is generated; if not, the raw material quality parameters are determined to be abnormal, and the initial raw material quality parameters are calculated in reverse. Construct a process quality transfer network, calculate the sensitivity matrix of the final product quality parameters to the process parameters, construct a joint optimization objective function to solve for the optimal process parameter adjustment amount, and execute joint control.

[0004] On the other hand, this application proposes an online monitoring system for cotton fiber quality based on intelligent sensors, used to implement the above-mentioned online monitoring method for cotton fiber quality based on intelligent sensors. The system includes: The data acquisition and calculation module is used to acquire the capacitance and loss tangent of the fiber under multi-band alternating voltage, and simultaneously acquire the fiber temperature, fiber layer thickness and conveying speed to form a detection record. The detection record is then calculated to obtain a local quality parameter vector. The fiber width detection and spatial distribution characterization module obtains local quality parameter vectors of multiple channels in the fiber width direction, constructs a width distribution vector and generates anomaly channel identification vectors, calculates width statistical features based on the width distribution vector, and outputs an outlet quality characterization data package. The process parameter acquisition and tracking module calculates the inlet quality parameter vector of the next process based on the outlet quality characterization data package of the previous process, obtains the process thermodynamic state and process operation parameters of each process, establishes a quality parameter evolution database, fits the process quality transfer function, and calculates the amount of quality deviation generated in each process. The anomaly location and inversion module performs reverse tracing of the process when the final product quality parameters deviate from the preset target parameters. It determines whether there is a process quality anomaly. If so, it generates an anomaly location result. If not, it determines that the raw material quality parameters are abnormal and inverts and calculates the initial raw material quality parameters. The joint optimization module constructs a process quality transfer network, calculates the sensitivity matrix of the final product quality parameters to the process parameters, constructs a joint optimization objective function to solve for the optimal process parameter adjustment amount, and executes joint control.

[0005] This invention provides a method and system for online monitoring of cotton fiber quality based on intelligent sensors. It has the following beneficial effects: 1. This invention establishes multiple detection channels along the fiber width direction, constructs a width distribution vector from the local quality parameter vectors of each channel according to the width position, and calculates the average width, standard deviation, coefficient of variation, and abnormal channel identifier. This transforms the microscopic local detection results into statistical features with macroscopic spatial semantics. The average width represents the overall fiber quality level, guiding the adjustment of global process parameters; the standard deviation and coefficient of variation represent the uniformity of quality along the width direction, used to assess equipment status; and the abnormal channel identifier precisely indicates the specific lateral location of quality deviation, guiding targeted adjustments of locally adjustable parameters. Utilizing a spatial scale conversion mechanism, the detection system can output decision-making basis with clear spatial orientation to the process control system, avoiding blind or erroneous adjustments caused by mismatch between detection and control scales. This improves the accuracy and efficiency of process adjustments, achieving a match between the spatial scale of detection results and the control scale of process parameters, thus enhancing the accuracy of process adjustments.

[0006] 2. This invention establishes a quality parameter evolution database containing inter-process quality transfer relationships and thermodynamic energy balance by simultaneously collecting fiber quality parameter vectors and process thermodynamic states at the inlet and outlet of each process. It then fits the process quality transfer function for each process and constructs a process quality transfer network with processes as nodes and fiber flow direction as edges. The sensitivity matrix of the final product quality parameters to the process parameters of each process is calculated using a perturbation method, and the globally optimal process parameter adjustment amount that minimizes the weighted quality deviation is then solved. Simultaneously, the initial raw material quality parameters are retrieved using accumulated thermal history data from multiple processes, achieving synergy between feedforward and feedback control. This allows process adjustments in each process to move from independent actions to global coordination. When the final product quality is abnormal, the process link causing the abnormality can be accurately located, and a joint optimization scheme and process parameter adjustment amount across processes can be provided. This avoids the problem of overall quality decline caused by local optimization of a single process, achieving global coordinated optimization of process parameters across multiple processes and improving the systematicness and effectiveness of the entire process quality control. Attached Figure Description

[0007] Figure 1 This is a flowchart illustrating the steps of the online monitoring method for cotton fiber quality based on intelligent sensors according to the present invention. Figure 2 This is a data transmission flowchart of the online monitoring method for cotton fiber quality based on intelligent sensors according to the present invention; Figure 3 This is an architecture diagram of the online monitoring system for cotton fiber quality based on intelligent sensors according to the present invention. Detailed Implementation

[0008] 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 some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0009] like Figures 1 to 2 As shown, the online monitoring method for cotton fiber quality based on intelligent sensors includes the following methods: Step S100 involves collecting the capacitance and loss tangent values ​​of the transported fibers under multi-band alternating voltage, and simultaneously collecting fiber temperature, fiber layer thickness, and transport speed to form a detection record. The detection record is then processed to obtain a local quality parameter vector. The detection record reflects the comprehensive electrical response characteristics of the fiber under specific electromagnetic field excitation and its physical state during transport. The original detection record is processed into a set of microscopic quality indicators with clear physical meaning using the local quality parameter vector, reflecting the intrinsic quality attributes of the fiber at the detection point, such as moisture content, maturity, neps content, and impurity content.

[0010] Step S200: Obtain local quality parameter vectors for multiple channels along the fiber width direction, construct a width distribution vector, generate anomaly channel identifier vectors, calculate width statistical features based on the width distribution vector, and output the exit quality characterization data package. By calculating the width statistical features, the detection results of multiple discrete channels along the fiber width direction are spatially aggregated and abstracted, thereby quantifying the uniformity and overall level of quality distribution along the width direction. The width statistical features reflect the quality of the fiber web's lateral consistency and the presence of skewed distribution. The exit quality characterization data package outputs a structured, comprehensive quality report containing spatial distribution information for the current process. It not only transmits the overall quality level to downstream processes but also accurately identifies the lateral location and statistical characteristics of anomalies, reflecting the spatial scale transformation of the detection results from point to surface.

[0011] Step S300: Based on the exit quality characterization data package of the previous process, calculate the inlet quality parameter vector of the next process, obtain the process thermodynamic state and process operating parameters of each process, establish a quality parameter evolution database, fit the process quality transfer function, and calculate the quality deviation generation of each process. The quality parameter evolution database is a historical record of the quality evolution of the entire process, reflecting the dynamic transfer law of quality in the processing chain as process operation and environmental conditions change. The process quality transfer function uses historical data from the quality parameter evolution database to establish a quantitative mapping relationship between the current process's exit quality and inlet quality, operating parameters, and thermodynamic state through regression analysis, reflecting the ability and effect of a specific process on the input quality. The quality deviation generation reflects the degree to which the process deviates from its normal transfer characteristics in actual operation and is the core criterion for locating abnormal processes.

[0012] Step S400: When the final product quality parameters deviate from the preset target parameters, reverse tracing of the process is performed to determine if there is a process quality anomaly. If so, an anomaly location result is generated; otherwise, the raw material quality parameters are determined to be abnormal, and the initial raw material quality parameters are calculated in reverse. By generating the anomaly location result, when the final product quality is abnormal, the specific process step causing the quality problem can be quickly and accurately identified by tracing the quality deviation of each process in reverse, reflecting the spatial location and severity of the anomaly source. By calculating the initial raw material quality parameters in reverse, the true quality level that the raw material itself should possess after excluding the influence of process processing is reflected.

[0013] Step S500 involves constructing a process quality transfer network, calculating the sensitivity matrix of the final product quality parameters to the process parameters, constructing a joint optimization objective function to solve for the optimal process parameter adjustment amount, and executing joint control. By constructing the process quality transfer network, the quality transfer functions of each process and the material flow delay relationships between them are linked in a directed graph, forming a complete quality evolution model from raw materials to finished products, reflecting the causal chain and coupling relationship of quality formation throughout the entire process. The sensitivity matrix is ​​calculated by applying small perturbations to the adjustable process parameters of each process and calculating their impact gradient on the final product quality parameters, reflecting the control effectiveness and leverage of different processes and different process parameters on the final product quality. The optimal process parameter adjustment amount is obtained by solving an optimization problem with the goal of minimizing the weighted quality deviation, under the constraint of satisfying the adjustable range of each process parameter, to obtain a set of process parameter changes that make the final product quality closest to the target value, reflecting a collaborative control instruction from the perspective of global optimization rather than local optimization. By implementing joint control, combining feedforward control with feedback control, the optimal adjustment amount is dynamically distributed to the execution mechanisms of each process, reflecting the system's proactive prediction and rapid response capability to quality fluctuations, and realizing closed-loop, collaborative, and intelligent control of the entire process.

[0014] In this embodiment, the steps for calculating the local quality parameter vector from the detection records include: Step S101: Collect the capacitance and loss tangent of the conveyed fiber under multi-band alternating voltage, and simultaneously collect the fiber temperature, fiber layer thickness, and conveying speed to form a detection record. A non-contact capacitance measuring device is installed on the conveying path of the cotton spinning fiber. This device includes two sets of opposing measuring electrodes with a measuring gap between them. The conveyed fiber does not directly contact the electrodes as it passes through the measuring gap. By applying a multi-band alternating voltage to the measuring electrodes and measuring the complex impedance between the electrodes, the real part is calculated by a digital signal processor after analog-to-digital conversion as the capacitance value, and the imaginary part as the loss tangent value. Based on the thermoelectric effect of a thermistor or infrared temperature measuring element, the temperature measuring element is placed close to the surface of the conveyed fiber, and the voltage signal is continuously collected and converted into a temperature value to obtain the fiber temperature. Based on the time-of-flight of a laser displacement sensor or ultrasonic sensor, the distance change from the sensor probe to the fiber layer surface is measured, calibrated, and converted into a thickness value to obtain the fiber layer thickness. Based on the pulse count of a rotary encoder installed at the end of the conveying roller shaft, the cumulative number of pulses per unit time is calculated and multiplied by the linear displacement corresponding to each pulse to obtain the instantaneous speed as the conveying speed. The data acquisition card synchronously triggers each sensor, reads the values ​​of all sensors at the same sampling time, attaches an absolute time tag, and stores them in the two-dimensional table or structured data stream in the buffer according to the sampling time order to form a complete detection record. The detection record includes a structured set of capacitance value, loss tangent value, fiber temperature, fiber layer thickness and conveying speed collected at the same time stamp.

[0015] Step S102: Extract the capacitance value, loss tangent, and fiber temperature from the test record. Calculate the equivalent capacitance value at the standard temperature based on the capacitance value and a preset temperature correction factor. Multiply the measured capacitance value by the temperature correction factor to obtain the equivalent capacitance value at the standard temperature. The temperature correction factor is composed of the product of the temperature coefficient of the fiber dielectric constant and the difference between the measured temperature and the standard temperature, plus one. By calculating the equivalent capacitance value, the influence of fiber temperature changes on the capacitance measurement is eliminated. This allows capacitance values ​​measured at different ambient temperatures to be unified on a standard temperature reference for comparison and subsequent calculations, thereby ensuring that the accuracy of calculations for quality parameters such as moisture content and maturity is not affected by temperature drift. The temperature correction factor is obtained from the correction coefficient table that has been pre-calibrated experimentally and stored in the system memory. It is set according to the inherent physical property of the dielectric constant of the fiber material changing with temperature. For a specific type of cotton fiber, the rate of change of its dielectric constant at different temperatures is measured in a laboratory environment, i.e., the temperature coefficient. The temperature coefficient of the fiber dielectric constant is obtained from the fiber material physics handbook or a pre-conducted temperature-controlled dielectric spectrum experiment. The measured temperature is the fiber temperature value synchronously collected in step S101. The standard temperature is a reference temperature value set artificially for a unified benchmark, such as 20 degrees Celsius, and its value is stored in the system configuration parameters.

[0016] Step S103: Calculate fiber moisture content based on equivalent capacitance value in the low-frequency band; calculate fiber maturity coefficient based on equivalent capacitance value and loss tangent value in the mid-frequency band; analyze the generated pulse peaks based on the time series of equivalent capacitance value in the high-frequency band, and calculate cotton nip content index and impurity content index.

[0017] When calculating fiber moisture content, a characteristic frequency point is selected in the low-frequency band, and the ratio of the equivalent capacitance value to the capacitance value in the dry state is obtained as the capacitance ratio. The natural logarithm of this capacitance ratio is multiplied by the first calibration coefficient, the loss tangent is multiplied by the second calibration coefficient, and then the third calibration coefficient is added to obtain the fiber moisture content. Fiber moisture content refers to the percentage of water mass in the fiber relative to the fiber's dry mass, used to evaluate the fiber's regain state and reflect its dryness or hygroscopicity. The first, second, and third calibration coefficients are experimentally calibrated constants. During calibration, impedance spectra are measured at different moisture content levels, and the actual moisture content is determined using an offline oven method. Each calibration coefficient is then fitted to the data. The capacitance value in the dry state is the pre-measured capacitance value of the fiber in a completely dry state.

[0018] When calculating the fiber maturity coefficient, equivalent capacitance and loss tangent values ​​are collected at multiple frequency points in the mid-frequency band. The equivalent capacitance value versus frequency curve is plotted, and the average slope is calculated. Similarly, the loss tangent versus frequency curve is plotted, and the peak frequency and peak height are extracted. The average slope is multiplied by the fourth calibration coefficient, the peak frequency by the fifth calibration coefficient, and the peak height by the sixth calibration coefficient. These are then summed and added to the seventh calibration coefficient to obtain the fiber maturity coefficient. The fiber maturity coefficient is a dimensionless index characterizing the degree of development of cotton fiber cell wall thickness. It is used to evaluate the physical maturity of the fiber, reflecting its strength, bending, and dyeing properties. The fourth, fifth, and sixth calibration coefficients are experimentally calibrated. During calibration, cotton fiber samples of different maturity grades are used, and their impedance spectra are measured under the same temperature and density conditions. Simultaneously, the true maturity coefficient is determined using polarized light microscopy, and the parameters in the above formula are obtained by fitting the data.

[0019] When generating the nipple content index and impurity content index, in the time series of equivalent capacitance values ​​in the high-frequency band, the standard deviation and average value of the equivalent capacitance values ​​within a time window are calculated. The fluctuation index is obtained by dividing the standard deviation by the average value. An amplitude threshold is set, and events in which the equivalent capacitance value continuously exceeds the amplitude threshold in multiple consecutive sampling points within the window are recorded as pulse spikes. The number of pulse spikes is recorded, and the pulse density is obtained by dividing the number of pulse spikes by the product of the mass flow rate and the window duration. The comprehensive content index is obtained by multiplying the fluctuation index by the eighth weight and the pulse density by the ninth weight. For each pulse spike, the duration and the amplitude exceeding the amplitude threshold are measured. Pulses with long duration and moderate amplitude are classified as nipples, and pulses with short duration and large amplitude are classified as impurities. The densities of the two types of pulses are counted separately, and the outputs are the nipple content index and the impurity content index. The nipple content index is a relative value that quantitatively describes the number of entangled fiber knots (nipples) in the fiber web. It is used to evaluate the cleanliness of the fiber and reflects the mechanical damage or original impurity content of the fiber during processing. The impurity content index is a relative value that quantitatively describes the amount of non-fibrous inclusions (such as leaf debris and dust) in a fiber. It is used to evaluate the cleanliness of the fiber and reflects the purity level and impurity removal effect of the raw material.

[0020] In calculating pulse spikes, a pulse spike is further defined as an event in which the equivalent capacitance value deviates from the average value by more than a preset multiple standard deviation, such as 3 times the standard deviation, across multiple consecutive sampling points. Because cotton knots are relatively large and their dielectric constant is close to that of normal fibers, the resulting pulses have a longer duration but relatively smaller amplitude. Since pulses caused by impurities are typically shorter in duration and larger in amplitude, a classification threshold is set based on the pulse characteristics during the classification process. Cotton knot pulse counts and impurity pulse counts are then counted separately, and the cotton knot content index and impurity content index are output separately.

[0021] The low-frequency, mid-frequency, and high-frequency bands are defined based on the frequency range of the alternating voltage applied to the sensor. These bands are selected according to the ability of electric fields of different frequencies to penetrate the fiber layer and the different polarization relaxation characteristics of different components in the fiber (such as water molecules, cellulose, and impurities). The low-frequency band (e.g., 10kHz to 100kHz) is mainly used to sense the orientation polarization of water molecules, the mid-frequency band (e.g., 100kHz to 2MHz) is mainly used to reflect the dielectric relaxation of fiber cell walls, and the high-frequency band (e.g., 2MHz to 20MHz) is sensitive to the density fluctuations of small foreign objects and fiber aggregates. By dividing the frequency bands, targeted decoupled measurements of different quality indicators such as moisture content, maturity, neps, and impurities can be achieved.

[0022] Step S104: The fiber moisture content, fiber maturity coefficient, nipple content index, and impurity content index are integrated to obtain a local quality parameter vector. The local quality parameter vector includes the specific values ​​corresponding to the four components: fiber moisture content, fiber maturity coefficient, nipple content index, and impurity content index. It also includes the acquisition time and channel location identifier. This data is packaged into a complete local quality parameter vector data package for use in step S200.

[0023] In this embodiment, the process of outputting the export quality characterization data packet includes: Step S201: Obtain the local quality parameter vectors and their width position coordinates for each of the multiple channels in the fiber width direction, and summarize them to obtain the global quality parameter vector. Step S100's acquisition process is a channel acquisition process. Multiple detection channels are arranged at equal intervals in the fiber width direction. Each detection channel is assigned a unique number, and its physical distance from the left boundary is recorded as its width position coordinate. Each channel independently acquires and calculates its local quality parameter vector according to step S100, while simultaneously recording the width position coordinates of each channel, including the channel number and the corresponding lateral distance value. The local quality parameter vectors of all channels are summarized to form the global quality parameter vector. By summarizing and arranging the local quality parameter vectors of all channels according to their width position using the global quality parameter vector, a large data structure is formed that can completely describe the quality distribution across the entire fiber width. This reflects the overall lateral distribution of quality parameters across the entire fiber width at the same time, providing the data foundation for width uniformity analysis and anomaly area location.

[0024] Step S202: For each type of quality parameter in the global quality parameter vector, arrange them in order of width position coordinates to obtain the width distribution vector of that type of quality parameter. Each width distribution vector includes the quality parameter value and width position coordinates for each channel of the corresponding type. The global quality parameter vector includes four types of quality parameters: fiber moisture content, fiber maturity coefficient, neps content index, and impurity content index. The width distribution vector is a one-dimensional sequence formed by arranging the detection values ​​of each type of quality parameter in all channels in order of width position coordinates from left to right. It is used to visually present the changing trend and distribution pattern of a single quality index (such as moisture content) throughout the fiber width direction.

[0025] Step S203: Calculate the width mean, width standard deviation, and width coefficient of variation in the width distribution vector to obtain the width statistical characteristics. The width statistical characteristics are a set of summary statistical indicators obtained through mathematical calculations of the width distribution vector. This compresses the complex width distribution curve into several concise values ​​with clear physical meaning, thereby quickly evaluating the overall level and uniformity of fiber quality in the transverse direction. The width statistical characteristics include three components: width mean, width standard deviation, and width coefficient of variation. When generating the width statistical characteristics, first, obtain the complete width distribution vector for a certain quality parameter. Then, calculate the arithmetic mean of all values ​​in the vector as the width mean. Calculate the square root of the sum of the squares of the differences between each value and the mean to obtain the width standard deviation. Divide the width standard deviation by the width mean and multiply by 100% to obtain the width coefficient of variation. The width mean is used to characterize the overall level of this quality parameter, reflecting the overall quality of the fiber. A benchmark value representing the overall level is obtained by calculating the width mean. Width standard deviation refers to the average deviation of the detection values ​​from the average value of each channel. It is used to measure the absolute dispersion of quality in the width direction, reflecting the quality of fiber lateral consistency. The larger the standard deviation, the worse the uniformity. The fluctuation range is quantified by calculating the standard deviation. Width coefficient of variation is used to measure the relative dispersion of quality. It eliminates the influence of the average level on the evaluation of dispersion, making it convenient to compare uniformity between different processes and batches. The uniformity comparison across batches and processes can be achieved by calculating the coefficient of variation.

[0026] Step S204: Determine whether the quality parameter value of each channel is within the preset range. If not, assign a value of one and generate an abnormal channel identifier vector for that channel. If yes, assign a value of zero and determine that the quality parameter is qualified. The preset range is a normal value range set for each quality parameter (e.g., moisture content). Only when the detected value falls within this range is it considered qualified. It is set based on the process standards of this process, the quality requirements of raw materials for subsequent processes, and the empirical range derived from historical normal production data. The preset range is obtained from the process quality parameter standard configuration file pre-entered and stored by the process engineer, serving as the criterion for determining whether there is a quality abnormality in each channel. The abnormal channel identifier vector is a binary flag sequence corresponding one-to-one with the width position coordinates. It is used to accurately mark which horizontal channel positions have quality parameters exceeding the qualified range. Each vector element includes a channel number and an abnormal flag bit. An abnormal flag bit of "1" indicates that the channel is abnormal, and "0" indicates normal. The flag bits of all channels are arranged in order of width position to form the abnormal channel identifier vector.

[0027] Step S205: The width statistical characteristic, the anomaly channel identifier vector, and the local quality parameters of each channel are combined into an export quality characterization data package, and the width average value is used as the overall export quality parameter vector for this process. The export quality characterization data package is the final quality report output by this process, containing rich spatial statistical information. It is used to provide structured data containing both overall quality level, lateral uniformity, and precise location of anomalies to the host computer control system or the next process. The overall export quality parameter vector refers to a parameter that can represent the overall quality level of the exported fiber of this process with a single value. As an input and output variable for quality transfer between processes, it simplifies the complexity of the process quality transfer function, allowing the quality transfer between different processes to be described by a scalar rather than a distribution vector. In this embodiment, since the width average value represents the overall level of the output quality of this process, while uniformity information such as width standard deviation and width coefficient of variation, although important, is already included in the export quality characterization data package for anomaly diagnosis and local control, using the width average value as a state variable in simplifying the macroscopic transfer model between processes captures the main contradiction and avoids an overly complex model.

[0028] In this embodiment, the steps for calculating the inlet quality parameter vector of the next process based on the outlet quality characterization data package of the previous process, and obtaining the process thermodynamic state and process operating parameters of each process, include: Step S301: Set the inlet quality parameter vector of the first process to the offline detection value. Since no upstream process provides online detection data before the first process, when the production line starts or the raw material batch is changed, the operator takes samples from the fiber bales and sends them to the laboratory to use standard testing equipment to determine the moisture content, maturity coefficient, neps content index, and impurity content index of that batch of raw materials. These measured values ​​are manually entered into the system's first process inlet parameter configuration interface, and the system reads and sets them as the inlet quality parameter vector of the first process. The offline detection value refers to the raw material quality indicators measured using standard methods in a laboratory environment, not on the production line.

[0029] For each process following the first process, the conveying distance between the exit of the previous process and the entrance of the current process is measured. The conveying distance is divided by the current fiber conveying speed to obtain the transmission delay time. Based on the conveying speed, the transmission delay time between adjacent processes is calculated. From the exit quality characterization data package of the previous process, the overall exit quality parameter vector corresponding to the moment after subtracting the transmission delay time is extracted and used as the entrance quality parameter vector of the current process at the current moment. By introducing transmission delay time compensation, the quality data of upstream and downstream processes are correctly correlated on the time axis, thereby establishing an accurate quality transfer relationship between processes. The entrance quality parameter vector is the set of intrinsic quality state parameters carried by the fiber when entering a certain process. It serves as the input variable of the quality transfer function of that process and is used to predict the exit quality of that process. It includes four components: fiber moisture content, maturity coefficient, nipple content index, and impurity content index at that moment.

[0030] Step S302: Fiber temperature, ambient temperature, and relative humidity are collected at the inlet and outlet of each process as the process thermodynamic state. Process operating parameters are obtained from the process control system. Process operating parameters include draw ratio, pressure, and operating speed. The changes in fiber temperature and moisture content between the inlet and outlet of each process are calculated. The process thermodynamic state refers to a comprehensive description of the environmental and internal thermodynamic conditions affecting fiber processing quality and dielectric properties. It provides an environmental correction factor for the process quality transfer function because thermodynamic conditions such as temperature and humidity affect the fiber's moisture content and mechanical properties, thus altering the process's effect on the fiber. The process thermodynamic state includes four parameters: inlet fiber temperature, outlet fiber temperature, ambient temperature, and relative humidity. Data is collected by installing temperature sensors at the inlet and outlet of each process, and temperature and humidity sensors inside or adjacent to the process equipment. These sensor readings are triggered simultaneously with the collection of quality parameters, and the four readings are packaged into a single thermodynamic state data record. Fiber temperature reflects the thermal state of the fiber itself; process ambient temperature reflects the background temperature of the processing environment and is used to analyze heat exchange; relative humidity reflects the water vapor content in the air and is used to predict the moisture absorption or release trend of the fiber.

[0031] Process operation parameters refer to the key adjustable operating parameters of the processing equipment in the current process. As input variables of the process quality transfer function, they reflect the impact of equipment settings on quality evolution. These parameters include three parameters: draw ratio, pressure, and running speed. The system communicates with the programmable logic controller (PLC) of the process equipment through an industrial fieldbus, periodically reads the currently set draw ratio, pressure value of the pressure roller, and speed setting value of the drive roller, and writes them into the process operation parameter record.

[0032] The fiber temperature change is the difference between the fiber temperature at the outlet and the fiber temperature at the inlet of the process, providing an energy balance index for the process quality transfer function that reflects the degree of heat absorption or release by the fiber in this process. The moisture content change is the difference between the fiber moisture content at the outlet and the fiber moisture content at the inlet of the process, reflecting the amount of humidity change that occurs in the fiber during this process due to mechanical action, heat exchange, or environmental exchange.

[0033] In this embodiment, the implementation steps for establishing a quality parameter evolution database, fitting the process quality transfer function, and calculating the quality deviation of each process include: Step S303: Based on the inlet quality parameter vector, outlet overall quality parameter vector, process thermodynamic state, and process operating parameters of all processes, a quality parameter evolution database is constructed. This database includes records generated for each process in each inspection cycle. Each record contains a process identifier, timestamp, inlet quality parameter vector, outlet overall quality parameter vector, inlet fiber temperature, outlet fiber temperature, inlet fiber moisture content, outlet fiber moisture content, fiber temperature change, moisture content change, process ambient temperature, process ambient relative humidity, and process operating parameters. The quality parameter evolution database is a continuously growing historical dataset stored in the form of relational tables on the hard drive or solid-state drive of an industrial computer, providing sufficient historical sample data with time and operating condition labels for fitting the process quality transfer function.

[0034] Step S304: Establish a process quality transfer function for each process based on database records and store it in the quality parameter evolution database. The process quality transfer function is a linear regression model established for each process based on historical database records accumulated in the quality parameter evolution database. The model uses the outlet quality parameter as the dependent variable and the inlet quality parameter vector, process operating parameters, environmental conditions, fiber temperature change, and moisture content change as independent variables. Specifically, it is a multiple linear regression equation that can predict the output outlet quality based on the current process's input conditions and operating settings. When establishing the process quality transfer function, once the number of records accumulated for that process in the quality parameter evolution database reaches a preset sample size threshold (e.g., one thousand records), the system calls the least squares regression program in the quality parameter evolution database to extract the corresponding category independent and dependent variables. The system then solves for the coefficients of each independent variable in the linear regression equation, minimizing the sum of squares of the differences between the predicted and actual values ​​for all samples. The resulting combination of coefficients is the quality transfer function for that process, and it is stored in the storage area associated with that process in the quality parameter evolution database.

[0035] Step S305: Substitute the obtained inlet quality parameter vector, process operating parameters, and process thermodynamic state of the current process into the process quality transfer function to calculate the predicted outlet quality value. The predicted outlet quality value is the theoretical expected value of the outlet quality of the process calculated by substituting the actual inlet quality parameters, process operating parameters, and thermodynamic state of the current process into the process quality transfer function fitted in step S304. It includes four components: predicted outlet moisture content, predicted outlet maturity coefficient, predicted outlet neps content index, and predicted outlet impurity content index. It is used to compare with the subsequently measured outlet quality parameters to determine whether the process has introduced additional quality deviations.

[0036] Step S306: Obtain the measured overall quality parameter vector of the outlet, calculate the difference between the overall quality parameter vector of the outlet and the predicted value of the outlet quality, and obtain the quality deviation amount generated by the process itself. First, based on step S205, obtain the measured overall quality parameter vector of the current process. Then, obtain the predicted value vector of the outlet quality calculated in step S305. Then, subtract the corresponding component of the predicted value vector from each component of the measured vector. That is, subtract the predicted outlet moisture content from the measured outlet moisture content to obtain the moisture content deviation, subtract the predicted outlet maturity coefficient from the measured outlet maturity coefficient to obtain the maturity coefficient deviation, and so on. Combine the four deviation values ​​obtained to form the quality deviation amount vector. The quality deviation amount refers to the abnormal quality component introduced by the process itself due to unexpected disturbances (such as mechanical vibration, local blockage, sensor drift, etc.). Specifically, it includes four components: moisture content deviation, maturity coefficient deviation, cotton nip content index deviation, and impurity content index deviation. The magnitude of the value quantifies the degree to which the process deviates from its normal quality transmission function and is the core criterion for locating the source of quality anomalies.

[0037] In this embodiment, the steps for reverse tracing of the process when the final product quality parameters deviate from the preset target parameters include: Step S401: Obtain the final product quality parameters and compare them with the preset target parameters. When the final product quality parameters deviate from the preset target parameters, conduct reverse traceability of the process. Compare each quality parameter of the obtained final product quality parameters with the preset target parameters. If all four parameters fall within their respective preset target parameter ranges, it is determined that the product quality is qualified and the process ends; if any one parameter falls outside the preset target range, it is determined that the product quality parameters deviate from the preset target parameters, triggering the reverse traceability process of step S402. Among them, the final product quality parameters refer to the set of overall quality indicators of fibers or yarns at the outlet of the last process (such as the yarn forming process), which is the ultimate basis for judging whether the product quality of the entire production line is qualified; specifically, it includes four components: the moisture content of the final product, the fiber maturity coefficient, the neps content index, and the impurity content index, which are directly obtained from the overall quality parameter vector of the outlet quality characterization data packet at the outlet of the last process. The preset target parameters are the qualified ranges or target values set for the final product quality parameters, including the upper and lower limits of the moisture content of the final product, the lower limit of the maturity coefficient, the upper limit of the neps content index, the upper limit of the impurity content index, etc.; they are set based on the product quality grade standard, customer order requirements, and the minimum requirements for raw material quality in processing (such as weaving, dyeing); the preset target parameters are stored in the product process specification database and are pre-entered by the process engineer according to the product variety and quality grade; in actual use, it is generally set as the moisture content target value of combed cotton yarn to be 7.0% ± 0.5%, the maturity coefficient not less than 0.85, the neps content index not exceeding 80 per gram, and the impurity content index not exceeding 20 per gram, and the specific values vary with the product grade.

[0038] Step S402: Starting from the last process, obtain the measured overall quality parameter vector at the outlet and calculate the quality deviation generation amount based on the process quality transfer function. Take the absolute value of the quality deviation generation amount and compare it with the preset process deviation threshold: If the quality deviation generation amount exceeds the process deviation threshold, it means that significant quality anomalies have been introduced in this process, reflecting problems such as mechanical failures, parameter drifts, local blockages, or environmental mutations in this process, and it is necessary to stop the machine for maintenance or adjustment. Then, determine that this process is a quality abnormal link, output the process identification and the quality deviation generation amount as the abnormal positioning result, and stop the traceability; the abnormal positioning result includes the identification of the abnormal process, the moment when the abnormality occurs, the specific component in the quality deviation generation amount that exceeds the threshold, and the amplitude that exceeds the threshold, which is used to guide the maintenance personnel to quickly locate the problem process and the problem type and reduce the downtime for troubleshooting.

[0039] If the amount of quality deviation does not exceed the process deviation threshold and the process is not the first process, it means that the process itself is operating normally and has not introduced unacceptable additional quality fluctuations. This reflects that the equipment status, process parameters and environmental control of the process are under control. In this case, the inlet quality parameter vector of the process is used as the overall outlet quality parameter vector measured by the previous process, and the previous process is traced for judgment. If the amount of quality deviation does not exceed the process deviation threshold, and the process is the first process, then the raw material quality parameters are determined to be abnormal.

[0040] The process deviation threshold is an acceptable maximum absolute value set for each component of the quality deviation generated in each process (moisture content deviation, maturity deviation, nipple deviation, and impurity deviation). It is based on the standard deviation multiple (usually three times the standard deviation) of the statistical distribution of quality deviation generated in the process under normal and stable production conditions, or an empirical value set by the process engineer based on experience. The process deviation threshold is stored in the process parameter configuration file and serves as a threshold value for judging whether the process is a quality abnormality link. Generally, it is set as follows for the carding process: moisture content deviation threshold ±0.3%, maturity deviation threshold ±0.05%, nipple deviation threshold ±15 pieces / gram, and impurity deviation threshold ±5 pieces / gram. For the drawing process, due to the self-regulating and leveling effect, the process deviation threshold is usually set more strictly.

[0041] By using the inlet quality parameter vector of this process as the overall outlet quality parameter vector of the previous process, the aim is to indirectly calculate the quality of the previous process's outlet by utilizing the material conservation and mass transfer relationships between processes when it is impossible to directly measure the outlet quality of the previous process. This solves the problem of not being able to install online detection sensors due to physical space limitations, such as a long-distance conveying pipeline or multiple merging points between the outlet of the previous process and the inlet of the current process.

[0042] In this embodiment, the process of inverting and calculating the initial raw material quality parameters includes: Step S403: When raw material quality parameters are determined to be abnormal, starting from the first process, based on the quality transfer function of each process, the overall quality parameter vector of the exit of each process is sequentially expressed as a composite function of the initial raw material quality parameters. Abnormal raw material quality parameters include one or more of the following: moisture content, maturity coefficient, neps content index, or impurity content index, exceeding the preset raw material acceptance standard. This is used to clarify that the root cause of the problem lies in the purchased raw materials rather than the production and processing process, providing a basis for claiming compensation from the supplier or adjusting the raw material ratio.

[0043] When the raw material quality parameters are determined to be abnormal, starting from the first process, the exit quality parameters of each process are sequentially expressed as a composite function of the initial raw material quality parameters. The composite function is: That is, the quality of the first process exit is equal to the calculation result of the quality transfer function of the first process with the initial raw material quality parameters as input; The quality at the exit of the second process is equal to the result of the quality transfer function of the second process calculated with the quality at the exit of the first process as input; this process continues until the last process, resulting in a composite function relating to the initial raw material quality parameters. By using the output of the quality transfer function of the first process as the input of the quality transfer function of the second process, and the output of the quality transfer function of the second process as the input of the quality transfer function of the third process, and so on, the final product quality is ultimately expressed as a composite function relating to the initial raw material quality parameters. The resulting composite function is a chain of nested functions relating the final product quality to the initial raw material quality, formed by connecting the quality transfer functions of multiple processes in the direction of material flow.

[0044] The initial raw material quality parameters are obtained by substituting the measured overall quality parameter vectors, process thermodynamic states, and process operating parameters of each process into a composite function. When inverting the initial raw material quality parameters, the overall quality parameter vectors, process operating parameters, and process thermodynamic states of all processes are first obtained from the measured data of each process. These known quantities are substituted into the composite function equation system. At this point, only the initial raw material quality parameters remain unknowns in the equation system, and the number of equations (i.e., the measured overall quality parameter vectors of each process) exceeds the number of unknowns. The least squares estimation algorithm in the numerical computation library is then used to find a set of initial raw material quality parameters that minimizes the sum of the squares of the differences between the predicted and measured export qualities of each process. This set of parameter values ​​is the inverted initial raw material quality parameter. The initial raw material quality parameter is the intrinsic quality state of the raw cotton fiber before entering the first process, including four components: initial moisture content, initial maturity coefficient, initial neps content index, and initial impurity content index. The specific values ​​of each mass component of the initial raw material are obtained by solving a system of composite function equations, providing a quantitative basis for raw material grading and cotton blending adjustments.

[0045] In this embodiment, the steps for constructing a process quality transfer network and calculating the sensitivity matrix of the final product quality parameters to the process parameters include: Step S501: Using each process as a node and the flow direction of the conveying fiber between processes as directed edges, a process quality transfer network is constructed. Each node is associated with the process quality transfer function of that process and the quality deviation generation of the process itself, and each edge is associated with the transmission delay time. First, the process list and physical connection relationships between processes defined in the production line configuration file are read. Then, a node object is created for each process, and the process quality transfer function coefficient obtained from step S304 and the quality deviation generation obtained from step S306 are assigned to the node. A directed edge is created between each pair of adjacent processes, and the transmission delay time calculated from step S301 is assigned to the edge. Finally, all nodes and edges are connected according to the actual process sequence and stored in a graph data structure in memory. Each process node, by associating with the process quality transfer function, simulates the processing state of the process for the input quality parameters, and by associating with the quality deviation generation, reflects the current abnormal state of the process. Each edge, by associating with the transmission delay time, achieves spatiotemporal alignment of upstream and downstream process quality data. The process quality transfer network completely describes the transmission path and evolution relationship of quality parameters throughout the entire process.

[0046] Step S502: Keeping the process parameters of other processes unchanged, calculate the partial derivative of the final product quality parameter with respect to each process parameter of the process to obtain the sensitivity of each adjustable process parameter of each process, forming a sensitivity matrix. The rows of the sensitivity matrix correspond to the process parameters, and the columns correspond to the final quality parameters. When generating the sensitivity matrix, first copy the complete state of the current process quality transfer network as a baseline, and then perform the following operations on each adjustable process parameter in sequence: add a small perturbation to the parameter based on its current value (e.g., increase the draw ratio by 0.01), keep all other process parameters unchanged, substitute the perturbed parameter into the process quality transfer network, and calculate the exit quality of each process sequentially from the first process until the final exit product quality parameter is calculated. Subtract the final product quality parameter in the baseline state from the calculated final product quality parameter, and then divide by the perturbation amount to obtain an approximate value of the partial derivative of the process parameter with respect to the final product quality. Repeat the above operation for all adjustable process parameters and all final product quality components, and fill the calculated partial derivatives into the corresponding positions of the matrix to form a complete sensitivity matrix. The resulting sensitivity matrix is ​​a two-dimensional numerical table. Its rows correspond to the adjustable process parameters in each process, such as the draft ratio of the first drawing process and the pressure of the second drawing process. Its columns correspond to the components of the final product quality parameters, including the final moisture content, the final maturity coefficient, the final neps content index, and the final impurity content index. Each element in the sensitivity matrix represents how much the final product quality parameter will change when a certain process parameter changes by a tiny unit, i.e., the approximate value of the partial derivative. It is used to quantitatively reveal the strength and direction of the influence of different process parameters in different processes on the final product quality, providing gradient direction information for optimization and control. The sensitivity matrix is ​​stored in computer memory in the form of a two-dimensional floating-point array.

[0047] In this embodiment, the steps of constructing a joint optimization objective function to solve for the optimal process parameter adjustment and executing joint control include: Step S503: Determine the set of final product quality parameters that need to be optimized simultaneously, and set target values ​​and priority weights for each final product quality parameter. Based on the quality comparison results of step S401, only those parameters whose actual test values ​​deviate from the preset target range need to be included in the optimization set, forming the set of final product quality parameters that need to be optimized. The set of final product quality parameters that need to be optimized refers to those quality indicators that are currently unqualified in actual testing and need to be corrected by adjusting process parameters. Specifically, it includes the name of the unqualified component and its current measured value, such as the current value of the cotton nip content index of 95 pieces / gram, which is used to clarify the optimization target and avoid unnecessary adjustments to indicators that have already passed. Meanwhile, by setting target values ​​and priority weights for each final product quality parameter, the quality control requirements are quantified into numerical targets that the optimization algorithm can handle. Priority weights are used to weigh trade-offs when multiple quality indicators conflict (e.g., reducing neps may increase impurities). The target values ​​for each quality component are read from the product process specification database (e.g., nep target value ≤ 60 pieces / gram), and the priority weight coefficients preset by the process engineer are read and assigned to each parameter in the optimization target list.

[0048] Step S504: Construct a joint optimization objective function and solve for the optimal process parameter adjustment amount. The joint optimization objective function is a weighted sum of squares function, which multiplies the squares of the deviation values ​​(differences between actual and target values) of all the final product quality parameters to be optimized by their respective priority weights and then sums them to obtain a single scalar value. The joint optimization objective function specifically includes the deviation squared term of each optimization objective component, the corresponding weight coefficient, and the entire summation expression, which serves as the loss function of the optimization algorithm. The goal of the optimization process is to find a set of process parameter adjustment amounts that minimizes this function value.

[0049] When generating the optimal process parameter adjustment, the current process parameter values ​​are used as the initial point. The sensitivity matrix calculated in step S502 provides the gradient direction. An iterative search method (such as gradient descent) is employed. In each iteration, the gradient of the objective function relative to each process parameter is calculated. A step size is calculated along the gradient descent direction, and the process parameter adjustment is updated. Then, the process parameters are re-substituted into the process quality transfer network to calculate the new predicted final product quality value. The objective function value is recalculated, and this process is repeated until the objective function value is less than the preset convergence threshold or the maximum number of iterations is reached. The process parameter adjustment obtained at this point is the optimal solution. The optimal process parameter adjustment refers to the magnitude by which a set of process parameters for each process step should be changed, obtained by solving for the minimum value of the joint optimization objective function. Specifically, this includes the adjustment value corresponding to each adjustable process parameter, such as increasing the drafting ratio of the first drawing pass by 0.05 and the pressure of the second drawing pass by 2 Newtons. This provides specific and quantitative control instructions to directly guide production operations or automatic control systems.

[0050] Step S505, joint control includes feedforward control and feedback control. Joint control combines the predictive nature of feedforward control with the corrective nature of feedback control to achieve dual assurance of quality throughout the entire process; feedforward control pre-sets the parameters of each process based on the raw material quality before production begins to avoid batch non-conformities caused by raw material fluctuations; feedback control monitors quality deviations in real time during production and dynamically corrects parameters to compensate for residual fluctuations and sudden disturbances that feedforward control could not completely eliminate.

[0051] Feedforward control inputs initial raw material quality parameters into the process quality transfer network before production begins, pre-calculates and distributes the initial process parameters for each process. Feedforward control, based on the inverse calculation of the initial raw material quality parameters, pre-calculates the optimal process parameters for each process before batch production begins and distributes them to the equipment for execution. Specifically, it includes the initial raw material quality parameters, the quality transfer function for each process, and the calculated optimal process parameter setpoints, ensuring the equipment is in an optimal working state matching the quality of the batch of raw materials before they arrive, shortening the adjustment time after startup. The control process is as follows: after changing the raw material batch, the initial raw material quality parameters are first inverted in step S403, then these parameters are input into the process quality transfer network. The optimal process parameter adjustment amount is solved using the optimization objective function in step S504 (where the objective value is the standard product objective value). However, the adjustment amount at this time is an absolute setpoint relative to the standard process parameter baseline value, not an increment. These setpoints are sent to the programmable logic controllers (PLCs) of each process via industrial Ethernet. The PLCs complete the parameter switching before receiving the raw materials.

[0052] Feedback control periodically acquires the overall quality parameter vector of the output from each process during production. When the quality deviation exceeds the process deviation threshold, it recalculates the sensitivity matrix and solves for the optimal process parameter adjustment based on the current process parameters. This optimal adjustment is then superimposed on the current process parameters and sent to the actuators of each process through the output interface. Feedback control involves repeatedly executing steps S401 to S504 at fixed intervals (e.g., every 30 seconds) during production to dynamically calculate the optimal process parameter adjustment relative to the current process parameters and send it out for execution in real time. This includes the currently measured final product quality parameters, the process parameters of each process, and the calculated optimal adjustment, used to continuously track quality fluctuations, correct deviations promptly, and maintain long-term product quality stability. During feedback control, the latest data is collected when each control cycle is triggered, and step S401 is executed to determine whether there is a deviation. If there is a deviation, steps S402 to S404 are executed to locate the abnormality (if there is an abnormality, an alarm is triggered). At the same time, regardless of whether there is an abnormality, steps S501 to S504 are executed to calculate the optimal adjustment amount (increment) relative to the current process parameters. Then, this increment value is superimposed on the process parameters currently running in each process PLC. The PLC adjusts the actuators (such as frequency converters, servo motors, and pneumatic control valves) accordingly.

[0053] In summary, in practical use, step S100 can be implemented using low-level drivers and digital signal processing libraries written in C / C++ on an embedded real-time operating system (such as FreeRTOS, VxWorks) or a Windows / Linux environment running on an industrial computer. Step S200 can be implemented on an industrial computer using high-level languages ​​such as Python, C#, or Java in conjunction with a relational database (such as SQLite, PostgreSQL); the aggregation of data from multiple channels can be achieved using multi-threading technology for parallel reception; the construction and sorting of the amplitude distribution vector can utilize the built-in array sorting functions of the language; the calculation of amplitude statistical features can call ready-made functions in statistical libraries (such as Python's NumPy, C#'s Math.NET Numerics); and the generation of abnormal channel identifier vectors is achieved through simple loop comparison statements. Step S300 can be implemented on an industrial computer using SCADA system software (such as WinCC, InTouch) or an open-source time-series database (such as InfluxDB) in conjunction with Python scripts. Process operating parameters (drawing ratio, pressure, speed) are read from the PLC via OPCUA or Modbus protocol. Transmission delay time is calculated based on conveying distance and speed. The quality parameter evolution database can be stored using the time-series database's write interface. The process quality transfer function can be performed using the scikit-learn library in Python or the lm function in R for least-squares fitting. Step S400 can be implemented on an industrial computer using Python or MATLAB. The reverse tracing process is implemented by iterating through the process list and comparing deviations with thresholds. The inversion solution of the composite function can be achieved by calling the leastsq function in Python's scipy.optimize library or the linalg.lstsq function in NumPy. Step S500 can be implemented on an industrial computer using Python in conjunction with a numerical optimization library (such as scipy.optimize); the construction of the process quality transfer network can be implemented using graph data structures in object-oriented programming; the calculation of the sensitivity matrix can be implemented using the finite difference method; the solution of the joint optimization objective function can be implemented using the gradient descent method or the more efficient BFGS algorithm; the logic of feedforward and feedback control is implemented through timer events and OPCUA client write operations.

[0054] The entire system can be compiled and deployed through an integrated development environment (such as Visual Studio or PyCharm), run on an industrial-grade edge computing gateway or a high-performance industrial control computer, and interconnected with PLCs and sensor networks in each process via Ethernet.

[0055] like Figure 3 As shown, the online monitoring system for cotton fiber quality based on intelligent sensors includes: The data acquisition and calculation module is used to acquire the capacitance and loss tangent of the fiber under multi-band alternating voltage, and simultaneously acquire the fiber temperature, fiber layer thickness and conveying speed to form a detection record. The detection record is then calculated to obtain a local quality parameter vector. The fiber width detection and spatial distribution characterization module obtains local quality parameter vectors of multiple channels in the fiber width direction, constructs a width distribution vector and generates anomaly channel identification vectors, calculates width statistical features based on the width distribution vector, and outputs an outlet quality characterization data package. The process parameter acquisition and tracking module calculates the inlet quality parameter vector of the next process based on the outlet quality characterization data package of the previous process, obtains the process thermodynamic state and process operation parameters of each process, establishes a quality parameter evolution database, fits the process quality transfer function, and calculates the amount of quality deviation generated in each process. The anomaly location and inversion module performs reverse tracing of the process when the final product quality parameters deviate from the preset target parameters. It determines whether there is a process quality anomaly. If so, it generates an anomaly location result. If not, it determines that the raw material quality parameters are abnormal and inverts and calculates the initial raw material quality parameters. The joint optimization module constructs a process quality transfer network, calculates the sensitivity matrix of the final product quality parameters to the process parameters, constructs a joint optimization objective function to solve for the optimal process parameter adjustment amount, and executes joint control.

[0056] In this embodiment, multiple detection channels are set along the fiber width direction. The local quality parameter vectors of each channel are used to construct a width distribution vector according to the width position. The average width, standard deviation of width, coefficient of variation of width, and abnormal channel identifier are calculated, converting the microscopic local detection results into statistical features with macroscopic spatial semantics. The average width represents the overall level of fiber quality and is used to guide the adjustment of global process parameters (such as machine speed and total draft ratio). The standard deviation of width and coefficient of variation represent the uniformity of quality in the width direction and are used to evaluate the equipment status. The abnormal channel identifier precisely indicates the specific lateral position where the quality deviation occurs and is used to guide the targeted adjustment of local adjustable parameters. By using a spatial scale conversion mechanism, the detection system can output decision basis with clear spatial orientation to the process control system, avoiding blind or erroneous adjustments caused by the mismatch between detection and control scales. This improves the accuracy and efficiency of process adjustment, achieves the matching of the spatial scale of detection results with the control scale of process parameters, and enhances the accuracy of process adjustment.

[0057] This invention establishes a quality parameter evolution database containing inter-process quality transfer relationships and thermodynamic energy balance by simultaneously collecting fiber quality parameter vectors and process thermodynamic states at the inlet and outlet of each process. It then fits the process quality transfer function for each process and constructs a process quality transfer network with processes as nodes and fiber flow direction as edges. The sensitivity matrix of the final product quality parameters to the process parameters of each process is calculated using a perturbation method, and the globally optimal process parameter adjustment amount that minimizes the weighted quality deviation is then determined. Simultaneously, the initial raw material quality parameters are retrieved using accumulated thermal history data from multiple processes, achieving synergy between feedforward and feedback control. This transforms process adjustments from isolated actions to global coordination. When the final product quality is abnormal, the process link causing the abnormality can be accurately located, and a cross-process joint optimization scheme and process parameter adjustment amount can be provided. This avoids the problem of overall quality degradation caused by local optimization of a single process, achieving global coordinated optimization of process parameters across multiple processes and improving the systematicness and effectiveness of the entire process quality control.

[0058] This application also provides an electronic device. The electronic device may include one or more processors and one or more memories. The memories store computer-readable code, which, when executed by the one or more processors, can perform the online monitoring method and system for cotton fiber quality based on smart sensors as described above.

[0059] The methods and systems according to the embodiments of this application can also be implemented using the architecture of the electronic device shown in this application. The electronic device may include a bus, one or more CPUs, ROM, RAM, a communication port connected to a network, input / output, a hard disk, etc. The storage device in the electronic device, such as a ROM or hard disk, may store the online monitoring method and system for cotton fiber quality based on smart sensors provided in this application. Furthermore, the electronic device may also include a user interface. Of course, the architecture shown in this application is merely exemplary; when implementing different devices, one or more components in the electronic device shown in this application may be omitted according to actual needs.

[0060] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising a reference structure" does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0061] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for online monitoring of cotton fiber quality based on intelligent sensors, characterized in that, The method includes: The capacitance and loss tangent of the fiber under multi-band alternating voltage are collected, and the fiber temperature, fiber layer thickness and conveying speed are collected simultaneously to form a detection record. The local quality parameter vector is obtained by solving the detection record. Obtain local quality parameter vectors for multiple channels in the fiber width direction, construct a width distribution vector and generate anomaly channel identification vectors, calculate width statistical features based on the width distribution vector, and output an outlet quality characterization data packet; Based on the exit quality characterization data package of the previous process, calculate the inlet quality parameter vector of the next process, obtain the process thermodynamic state and process operating parameters of each process, establish a quality parameter evolution database, fit the process quality transfer function, and calculate the amount of quality deviation generated in each process. When the final product quality parameters deviate from the preset target parameters, reverse trace of the process is performed to determine whether there is a process quality abnormality. If so, an abnormality location result is generated; if not, the raw material quality parameters are determined to be abnormal, and the initial raw material quality parameters are calculated in reverse. Construct a process quality transfer network, calculate the sensitivity matrix of the final product quality parameters to the process parameters, construct a joint optimization objective function to solve for the optimal process parameter adjustment amount, and execute joint control.

2. The online monitoring method for cotton fiber quality based on intelligent sensors according to claim 1, characterized in that, The process of solving the detection records to obtain the local quality parameter vector includes: Extract the capacitance value, loss tangent, and fiber temperature from the test records, and calculate the equivalent capacitance value at the standard temperature based on the capacitance value and the preset temperature correction factor; In the low-frequency band, the fiber moisture content is calculated based on the equivalent capacitance value. In the mid-frequency band, the fiber maturity coefficient is calculated based on the equivalent capacitance value and the loss tangent. Based on the time series of the equivalent capacitance value, the generated pulse peaks are analyzed, and the nip content index and impurity content index are calculated. The local quality parameter vector is obtained by integrating fiber moisture content, fiber maturity coefficient, nipple content index and impurity content index.

3. The online monitoring method for cotton fiber quality based on intelligent sensors according to claim 1, characterized in that, The process of outputting the export quality characterization data packet includes: Record each local mass parameter vector and its width and position coordinates, and summarize them to obtain the global mass parameter vector; For each type of quality parameter in the global quality parameter vector, arrange them in order of swath width position coordinates to obtain the swath width distribution vector of that type of quality parameter. Each swath width distribution vector includes the quality parameter value and swath width position coordinates of each channel of the corresponding type. Calculate the average width, standard deviation of width, and coefficient of variation of width in the width distribution vector to obtain the statistical characteristics of width. Determine whether the quality parameter value of each channel is within the preset range. If not, generate an abnormal channel identifier vector for that channel. The width statistical characteristic, the abnormal channel identification vector, and the local quality parameters of each channel are combined into an export quality characterization data package, and the average width is used as the overall export quality parameter vector of this process.

4. The online monitoring method for cotton fiber quality based on intelligent sensors according to claim 3, characterized in that, The process calculates the inlet quality parameter vector for the next process based on the outlet quality characterization data package of the previous process, and obtains the process thermodynamic state and process operating parameters for each process, including: Set the inlet quality parameter vector of the first process to the offline detection value; For each process after the first process, the transmission delay time between adjacent processes is calculated based on the conveying speed. The overall quality parameter vector of the previous process at the time after subtracting the transmission delay time is extracted from the output quality characterization data packet of the previous process and used as the input quality parameter vector of the current process at the current time. Fiber temperature, ambient temperature, and relative humidity are collected at the inlet and outlet of each process to determine the thermodynamic state of the process, and process operating parameters are obtained from the process control system.

5. The online monitoring method for cotton fiber quality based on intelligent sensors according to claim 4, characterized in that, The process of establishing a quality parameter evolution database, fitting the process quality transfer function, and calculating the quality deviation generated in each process includes: Based on the inlet quality parameter vector, outlet overall quality parameter vector, process thermodynamic state, and process operating parameters of all processes, a quality parameter evolution database is constructed, which includes database records generated by all processes in each detection cycle. Based on the database records, a process quality transfer function is established for each process and stored in the quality parameter evolution database; Substitute the obtained inlet quality parameter vector, process operating parameters, and process thermodynamic state of the current process into the process quality transfer function to calculate the predicted outlet quality value. Obtain the measured overall export quality parameter vector, calculate the difference between the overall export quality parameter vector and the predicted export quality value, and obtain the amount of quality deviation generated by the process itself.

6. The online monitoring method for cotton fiber quality based on intelligent sensors according to claim 1, characterized in that, When the final product quality parameters deviate from the preset target parameters, the process reverse tracing is performed, including: Starting from the last process, the measured overall quality parameter vector of the outlet is obtained and the amount of quality deviation is calculated based on the process quality transfer function. The amount of quality deviation is then compared with the preset process deviation threshold. If the amount of quality deviation exceeds the process deviation threshold, the process is determined to be a quality abnormality step. The process identifier and the amount of quality deviation are output as the abnormality location result, and the tracing is stopped. If the amount of quality deviation does not exceed the process deviation threshold, and the process is not the first process, then the inlet quality parameter vector of the process is used as the overall outlet quality parameter vector measured by the previous process, and the process is traced back to the previous process for judgment. If the amount of quality deviation does not exceed the process deviation threshold, and the process is the first process, then the raw material quality parameters are determined to be abnormal.

7. The online monitoring method for cotton fiber quality based on intelligent sensors according to claim 1, characterized in that, The process of inverting and calculating the initial raw material quality parameters includes: When determining that the raw material quality parameters are abnormal, starting from the first process, based on the quality transfer function of each process, the overall quality parameter vector of the exit of each process is expressed as a composite function of the initial raw material quality parameters. Substitute the measured overall quality parameter vector of each process, the thermodynamic state of the process, and the operating parameters of the process into the composite function to obtain the initial raw material quality parameters.

8. The online monitoring method for cotton fiber quality based on intelligent sensors according to claim 1, characterized in that, The construction of the process quality transfer network, and the calculation of the sensitivity matrix of the final product quality parameters to the process parameters of the process, include: With each process as a node and the flow direction of the conveying fiber between processes as a directed edge, a process quality transfer network is constructed. Keeping the process parameters of other processes unchanged, calculate the partial derivative of the final product quality parameter with respect to each process parameter of the process, obtain the sensitivity of each adjustable process parameter of each process, and form a sensitivity matrix. The rows of the sensitivity matrix correspond to the process parameters, and the columns correspond to the final quality parameters.

9. The online monitoring method for cotton fiber quality based on intelligent sensors according to claim 8, characterized in that, The process of constructing a joint optimization objective function to solve for the optimal process parameter adjustment and executing joint control includes: Determine the set of final product quality parameters that need to be optimized simultaneously, and set target values ​​and priority weights for each final product quality parameter; Construct a joint optimization objective function and solve for the optimal process parameter adjustment amount; The joint control includes feedforward control and feedback control. Before production begins, the feedforward control substitutes the initial raw material quality parameters into the process quality transfer network, pre-calculates the initial process parameters of each process, and issues them out. Feedback control periodically acquires the overall quality parameter vector of the outlet measured in each process during the production process. When the amount of quality deviation exceeds the process deviation threshold, the sensitivity matrix is ​​recalculated and the optimal process parameter adjustment is solved based on the current process parameters. The optimal process parameter adjustment is then superimposed on the current process parameters and sent to the actuators of each process through the output interface.

10. An online monitoring system for cotton fiber quality based on intelligent sensors, characterized in that, The system includes: The data acquisition and calculation module is used to acquire and calculate the detection records of the conveyed fibers to obtain a local quality parameter vector; The width detection and spatial distribution characterization module is used to obtain local quality parameter vectors of multiple channels in the fiber width direction, construct width distribution vectors and generate abnormal channel identification vectors, calculate width statistical features based on width distribution vectors, and output outlet quality characterization data packages. The process parameter acquisition and tracking module calculates the inlet quality parameter vector of the next process based on the outlet quality characterization data package of the previous process, obtains the process thermodynamic state and process operation parameters of each process, establishes a quality parameter evolution database, fits the process quality transfer function, and calculates the amount of quality deviation generated in each process. The anomaly location and inversion module is used to perform reverse tracing of the process when the final product quality parameters deviate from the preset target parameters, determine whether there is a process quality anomaly, and invert and calculate the initial raw material quality parameters. The joint optimization module is used to construct a process quality transfer network, calculate the sensitivity matrix of the final product quality parameters to the process parameters, construct a joint optimization objective function to solve for the optimal process parameter adjustment amount, and execute joint control.

Images