Intelligent monitoring method and system for chenille yarn steaming fabric chemical fiber oiling agent

Abstract

The invention relates to the technical field of intelligent monitoring, and discloses an intelligent monitoring method and system for chenille steamed yarn fabric chemical fiber oil. The method comprises the following steps: carrying out multi-wavelength near infrared spectrum and infrared thermal imaging acquisition on the chenille yarn steaming fabric in a yarn steaming treatment process to obtain oil agent double-layer distribution monitoring data; performing spectrum-temperature conversion on the oiling agent double-layer distribution monitoring data to obtain an oiling agent multi-layer permeation feature vector; inputting the oil agent multi-layer penetration feature vector into an oil agent penetration time sequence identification network for state analysis to obtain a quantitative penetration depth index; and performing penetration and uniformity dual-task evaluation based on the quantitative penetration depth index, and generating an oil agent deep penetration process parameter combination of the chenille yarn steaming process. According to the method, the optimal process parameter combination can be automatically generated according to different chenille fabric characteristics, closed-loop control from monitoring data to process parameters is realized, and the accuracy and the stability of oiling agent treatment are improved.

Description

Technical Field

[0001] This application relates to the field of intelligent monitoring technology, and particularly to an intelligent monitoring method and system for chemical fiber oil agents in chenille yarn steaming fabrics. Background Art

[0002] The treatment quality of chemical fiber oil agents in chenille yarn steaming fabrics directly affects the hand feeling, antistatic property, color fastness and overall performance of the fabrics. Traditional oil agent monitoring methods mainly rely on off-line sampling detection and empirical regulation, and cannot monitor the penetration state of oil agents in fabrics in real time, especially it is difficult to distinguish between the two different distribution states of surface adhesion amount and internal penetration amount. When evaluating the effect of oil agents in chenille fabrics, the internal penetration amount has a higher indicative significance than the surface adhesion amount, because only the oil agents that penetrate into the fiber interior can play a lasting role, and the oil agents attached to the surface are likely to be lost in subsequent processing.

[0003] Existing technology monitoring means lack the ability of multi-parameter joint analysis and cannot cope with the complex variables in the chenille yarn steaming process. The relationships between process parameters such as steaming temperature, pressure, humidity and time and the oil agent penetration effect are difficult to accurately quantify, resulting in the setting of process parameters mainly relying on empirical judgment, with obvious hysteresis and uncertainty. Especially for chenille fabrics with different densities and different component ratios, there are significant differences in their optimal oil agent treatment parameters, while traditional monitoring methods cannot provide differentiated process parameter optimization schemes. Summary of the Invention

[0004] This application provides an intelligent monitoring method and system for chemical fiber oil agents in chenille yarn steaming fabrics. This application can automatically generate an optimal process parameter combination according to the characteristics of different chenille fabrics, realize the closed-loop control from monitoring data to process parameters, and improve the accuracy and stability of oil agent treatment.

[0005] In the first aspect, this application provides an intelligent monitoring method for chemical fiber oil agents in chenille yarn steaming fabrics, and the intelligent monitoring method for chemical fiber oil agents in chenille yarn steaming fabrics includes:

[0006] Performing multi-wavelength near-infrared spectroscopy and infrared thermal imaging acquisition on the chenille yarn steaming fabric during the steaming process to obtain oil agent double-layer distribution monitoring data;

[0007] Performing spectral-temperature transformation on the oil agent double-layer distribution monitoring data to obtain oil agent multi-layer penetration feature vectors;

[0008] Inputting the oil agent multi-layer penetration feature vectors into an oil agent penetration time series recognition network for state analysis to obtain a quantitative penetration depth index;

[0009] Perform the dual tasks of penetration and uniformity evaluation based on the quantitative penetration depth index, and generate the combination of deep penetration process parameters of the sizing agent for the chenille yarn steaming process.

[0010] In a second aspect, the present application provides an intelligent monitoring system for sizing agents of chenille yarn steaming fabrics. The intelligent monitoring system for sizing agents of chenille yarn steaming fabrics includes:

[0011] A collection module, configured to perform multi-wavelength near-infrared spectroscopy and infrared thermal imaging collection on the chenille yarn steaming fabric during the steaming process to obtain sizing agent double-layer distribution monitoring data;

[0012] A transformation module, configured to perform spectral-temperature transformation on the sizing agent double-layer distribution monitoring data to obtain sizing agent multi-layer penetration feature vectors;

[0013] A state analysis module, configured to input the sizing agent multi-layer penetration feature vectors into a sizing agent penetration time series recognition network for state analysis to obtain a quantitative penetration depth index;

[0014] A generation module, configured to perform the dual tasks of penetration and uniformity evaluation based on the quantitative penetration depth index, and generate the combination of deep penetration process parameters of the sizing agent for the chenille yarn steaming process.

[0015] In the technical solution provided by the present application, through the combination of multi-wavelength near-infrared spectroscopy and infrared thermal imaging technologies, the accurate discrimination monitoring of the surface adhesion amount and internal penetration amount of the sizing agent is realized, overcoming the technical bottleneck that the traditional method cannot distinguish the distribution levels of the sizing agent. The sizing agent multi-layer penetration feature vectors constructed by the spectral-temperature transformation method realize the comprehensive characterization of the sizing agent distribution states in the surface layer, shallow layer, and deep layer of the fabric. The sizing agent penetration time series recognition network analyzes the continuously collected feature data, can capture the dynamic changes of the sizing agent penetration process in real time, and effectively distinguish different penetration states. The dual-task evaluation model of penetration and uniformity realizes the parallel evaluation and comprehensive consideration of penetration depth and uniformity through a shared encoding layer and a feature interaction mechanism. The generated comprehensive sizing agent penetration quality index provides a quantitative basis for process parameter optimization. The parameter mapping relationship and fuzzy control rule base established based on this comprehensive index can automatically generate the optimal process parameter combination according to the characteristics of different chenille fabrics, realizing the closed-loop control from monitoring data to process parameters, improving the accuracy and stability of sizing agent treatment, and at the same time avoiding over-steaming of the yarn and waste of the sizing agent by precisely controlling the sizing agent penetration process. Description of the Drawings

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained based on these drawings.

[0017] Figure 1 It is a schematic diagram of an embodiment of the intelligent monitoring method for chemical fiber oil agent of chenille yarn steaming fabric in the embodiments of the present application;

[0018] Figure 2 It is a schematic diagram of an embodiment of the intelligent monitoring system for chemical fiber oil agent of chenille yarn steaming fabric in the embodiments of the present application. Detailed implementation manners

[0019] The embodiments of the present application provide an intelligent monitoring method and system for chemical fiber oil agent of chenille yarn steaming fabric. Terms such as "first", "second", "third", "fourth", etc. (if any) in the specification, claims and above-mentioned accompanying drawings of the present application are used to distinguish similar objects, and do not necessarily describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "comprising" or "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units does not necessarily limit to those clearly listed steps or units, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0020] For easy understanding, the following describes the specific process of the embodiments of the present application. Please refer to Figure 1 , an embodiment of the intelligent monitoring method for chemical fiber oil agent of chenille yarn steaming fabric in the embodiments of the present application includes:

[0021] Step S101: Perform multi-wavelength near-infrared spectroscopy and infrared thermal imaging acquisition on the chenille yarn steaming fabric during the yarn steaming process to obtain oil agent double-layer distribution monitoring data;

[0022] It can be understood that the execution subject of the present application can be an intelligent monitoring system for chemical fiber oil agent of chenille yarn steaming fabric, or a terminal or a server. Specifically, no limitation is made here. The embodiments of the present application take the server as the execution subject as an example for illustration.

[0023] Specifically, a multi-parameter acquisition system consisting of a multi-point near-infrared detection system and a temperature distribution monitoring system is constructed. The system installs and calibrates a near-infrared spectral sensor array and an infrared thermal imaging device to ensure the acquisition of high-quality raw data. During the installation process, the near-infrared spectral sensor array is arranged in a ring, covering all key areas of the chenille fabric during the yarn steaming process and maintaining an appropriate distance from the fabric surface. The infrared thermal imaging device is positioned according to the layout and monitoring of the yarn steaming equipment and installed at the center of the equipment to capture the temperature changes on the fabric surface. The acquired raw spectral data comes from the multi-point near-infrared detection system and can monitor the spectral changes of the fabric at different wavelengths during the yarn steaming process. To eliminate the influence of background noise and light scattering, the raw spectral data is subjected to baseline correction and scattering correction to obtain near-infrared spectral data reflecting the distribution of the oil agent on and inside the fabric surface, effectively capturing the absorption characteristic peaks of the oil agent. At the same time, the raw thermal imaging data captured by the infrared thermal imaging device provides the surface temperature distribution information of the fabric during the yarn steaming process. By calculating the temperature gradient of these thermal imaging data, infrared thermal imaging data characterizing the penetration depth of the oil agent is obtained. Since the temperature distribution is closely related to the penetration depth of the oil agent, the thermal imaging data can effectively reveal the distribution of the oil agent on the fabric surface and the degree of penetration into the interior. The temperature gradient calculation helps to quantify the penetration state of the oil agent at each level by analyzing the temperature changes at different depths of the fabric. The near-infrared spectral data and the infrared thermal imaging data are fused to obtain oil agent distribution monitoring data. The spectral data reflects the surface adhesion of the oil agent, while the thermal imaging data provides the penetration state of the oil agent inside the fabric. Through data fusion, the surface adhesion amount and the internal penetration amount of the oil agent in the fabric are accurately distinguished, forming oil agent double-layer distribution monitoring data with a high information density.

[0024] Step S102: Perform a spectral-temperature transformation on the oil agent double-layer distribution monitoring data to obtain an oil agent multi-layer penetration feature vector;

[0025] Specifically, standard normal variate transformation is performed on the near-infrared spectral data to eliminate the interference caused by the deviation and noise of the spectral signal, obtaining a set of corrected characteristic spectral data that contains important information related to the molecular structure of the sizing agent, providing the distribution state of the sizing agent on the fabric surface and the penetration situation within a certain depth. By performing characteristic peak identification and analysis on the characteristic spectral data, the intensity change data of the sizing agent characteristic peaks are identified. These characteristic peaks appear in specific wavelength regions, reflecting the distribution and change of the sizing agent at different levels. A temperature gradient model characterizing the heat conduction characteristics is constructed based on the infrared thermal imaging data in the sizing agent bilayer distribution monitoring data. The thermal imaging data provides the temperature distribution information on the fabric surface, and these temperature data can reflect the penetration depth of the sizing agent. By analyzing the temperature changes in different depth regions, a temperature gradient model describing the heat conduction process is constructed to characterize the temperature changes in different depth regions of the fabric and reveal the penetration behavior of the sizing agent at each level. Based on the constructed temperature gradient model, hierarchical division processing is performed on the chenille yarn steaming fabric, dividing the fabric into three regions: the surface layer, the shallow layer, and the deep layer. Each level represents the penetration state of the sizing agent in different depth regions. The surface layer is in the depth range of 0 - 0.5 mm, the shallow layer is in the depth range of 0.5 - 1.5 mm, and the deep layer exceeds 1.5 mm. Through the layering method, more detailed sizing agent distribution information is obtained, providing an independent evaluation of the penetration state of the sizing agent for each layer. The temperature change rate and spectral characteristics of each region help to determine the penetration depth of the sizing agent in that layer, thereby further analyzing the distribution of the sizing agent in the fabric. After completing the hierarchical division, by combining the characteristic peak intensity change data with the hierarchical data structure, the sizing agent distribution index for each layer is calculated. These sizing agent distribution indexes can reflect the penetration degree and uniformity of the sizing agent at different levels. The sizing agent distribution index of the surface layer is calculated through near-infrared reflectance spectroscopy, the sizing agent distribution index of the shallow layer is calculated by combining diffuse reflectance spectroscopy and temperature gradient data, and the sizing agent distribution index of the deep layer is calculated by combining the temperature decay curve and the second derivative of the spectrum. Indexes such as the temperature change rate and spectral standard deviation are calculated to further evaluate the distribution characteristics of the sizing agent in each layer, obtaining the target characteristic data. Principal component analysis is performed on the target characteristic data. By extracting the main characteristics and reducing redundant information, the multi-layer penetration characteristic vector of the sizing agent is obtained.

[0026] In this embodiment, the chenille yarn steaming fabric is hierarchically divided based on the temperature gradient model. The heat conduction characteristics of the fabric are analyzed through the temperature gradient model, and different physical structure layers of the fabric are determined using this analysis. The temperature gradient model can reflect the heat propagation characteristics in the fabric. By calculating the heat conduction rate at each spatial position point, the temperature decay curve function characterizing the heat transfer characteristics in the fabric is obtained. By analyzing the temperature decay curve function, the mutation points of the temperature decay rate are determined. These mutation points reflect the change in the heat conduction speed and appear at the junctions between the physical structure layers. By identifying these mutation points, critical values are set for the depth of each layer, thereby clearly distinguishing the depth ranges of different physical structure layers. For example, the demarcation points between the surface layer, the shallow layer, and the deep layer. The depth critical values are adaptively matched with the preset depth demarcation points. The depth critical values are compared with the preset depth demarcation points to obtain a correction coefficient for the current chenille yarn steaming fabric structure, thereby optimizing the temperature gradient model to better adapt to the specific fabric structure and oil penetration characteristics. Based on the correction coefficient, a three-dimensional reconstruction of the temperature gradient model is performed to obtain a three-dimensional temperature distribution model including spatial coordinates and temperature values. The interlayer temperature gradient of the three-dimensional temperature distribution model is calculated to obtain a temperature change rate matrix characterizing the oil penetration state of each layer, showing the rate of temperature change between different layers and reflecting the oil penetration process in each layer. Through the calculation of the temperature change rate, the penetration state of each layer is quantified. According to the temperature change rate matrix and the spectral second derivative data, a characteristic mapping relationship of a three-layer structure is constructed. This characteristic mapping relationship combines the oil penetration characteristics of the surface layer, the shallow layer, and the deep layer to form a complete hierarchical data structure. Through this data structure, the oil penetration state of each layer and their mutual relationship are evaluated.

[0027] Step S103: Input the multi-layer oil penetration feature vector into the oil penetration time series recognition network for state analysis to obtain a quantitative penetration depth index;

[0028] Specifically, the oil agent multi-layer penetration feature vector is input into the input layer of the oil agent penetration time series recognition network. The feature vector contains the distribution information of the oil agent at different depth levels, including spectral data, temperature change data, and other characteristic values representing the penetration state of the oil agent. At the input layer, the network receives these multi-dimensional features and converts them into a series of feature sequence data streams. The oil agent penetration time series recognition network consists of multiple layers, including the first layer of LSTM (Long Short-Term Memory) hidden layer, the second layer of LSTM hidden layer, and the output layer. In the first layer of LSTM hidden layer, the feature sequence data stream undergoes the process of forward information extraction and backward information extraction. The LSTM network has the ability to memorize and can effectively capture the long-term and short-term dependencies in the data when processing time series data. In forward information extraction, the LSTM calculates the current hidden state based on the current input feature and the state information at the previous time point; in the backward information extraction process, the LSTM further processes the information transmitted from subsequent time points using the backpropagation algorithm. Through these two information extraction methods, the LSTM network extracts intermediate feature vectors containing the state of the gating unit from the time series data. These intermediate feature vectors contain the information of the current moment and integrate the memory of historical time points, which helps to accurately identify the dynamic changes in the oil agent penetration process. The intermediate feature vectors are passed to the second layer of LSTM hidden layer for processing. In the second layer of LSTM hidden layer, the network will perform advanced feature extraction on the intermediate feature vectors to identify more complex dynamic patterns in the oil agent penetration process. Through this processing, the network obtains the time series state features representing the oil agent penetration process. These state features reflect the real-time progress of the oil agent penetration and quantify the different stages of the penetration process. The output data of this layer is mapped to a three-dimensional state space. Based on the output values obtained from the three-dimensional state space, the oil agent penetration state vector is calculated, which includes the probability distribution of the oil agent surface attachment state, the penetration progress state, and the penetration completion state. When the probability value of the penetration completion state exceeds the preset target value, the network determines that the oil agent has completed penetration. This determination process can effectively distinguish the different stages of the oil agent penetration. For example, when the probability value of the penetration completion state is greater than the preset target value, it is considered that the oil agent has penetrated into the interior of the fabric and the penetration process has entered the completion stage. In the output layer, based on the feature vector corresponding to the oil agent penetration completion state, the effective penetration amount of the oil agent is calculated, and a quantitative penetration depth index representing the distribution ratio of the oil agent in each layer of the chenille yarn steaming fabric is obtained.

[0029] Step S104: Based on the quantitative penetration depth index, perform a dual-task evaluation of penetration and uniformity to generate a combination of oil agent deep penetration process parameters for the chenille yarn steaming process.

[0030] Specifically, multi-point sampling is performed on the chenille yarn steaming fabric based on the quantitative penetration depth index to obtain the penetration depth distribution data in the width direction of the fabric. Through multi-point sampling, the penetration depth data at different positions in the width direction of the fabric are obtained, reflecting the penetration degree of the sizing agent in the fabric and its distribution uniformity. The quantitative penetration depth index is input into the shared encoding layer of the dual-task neural network, and the input multi-dimensional data is transformed into intermediate feature representations, which contain key information about the penetration state and uniformity. By processing in this layer, the network extracts the global information during the penetration process. At this stage, the neural network comprehensively considers the mutual relationships of various input features. To enhance the interaction between tasks, the feature interaction module enhances the intermediate feature representations by introducing an attention mechanism, enabling the neural network to automatically focus on the key information in the penetration state evaluation and uniformity evaluation. Through the attention mechanism, the network assigns different weights according to the different importance of the data, making the information interaction between tasks more efficient and accurate. The enhanced features better describe the penetration state and distribution uniformity of the sizing agent in the fabric. The enhanced features are respectively input into the penetration state decoding layer and the uniformity decoding layer for dual decoding processing. The penetration state decoding layer calculates the penetration state evaluation value according to the input features, reflecting different stages of the sizing agent penetration process, while the uniformity decoding layer evaluates the uniformity of the sizing agent penetration according to the input features. These two decoding layers respectively focus on the two key factors of penetration depth and uniformity, and on this basis, evaluate the penetration quality of the sizing agent. Through the decoding process, the network can respectively obtain the penetration state and uniformity evaluation values. According to the penetration state evaluation value, the overall effective penetration amount of the fabric is calculated, reflecting the actual penetration degree of the sizing agent in the fabric. This value is obtained through the comprehensive calculation of the penetration state evaluation value, quantifying the penetration effect of the sizing agent. At the same time, according to the uniformity evaluation value, the coefficient of variation of the penetration amount in the width direction of the fabric is calculated. The coefficient of variation is an important index to measure the penetration uniformity, describing the uniformity of the sizing agent penetration at different width positions of the fabric. By calculating the coefficient of variation, it is analyzed whether the distribution of the sizing agent in the fabric is uniform. By combining the effective penetration amount and the coefficient of variation, a comprehensive index of the sizing agent penetration quality is obtained to evaluate the performance of the chenille yarn steaming fabric during the sizing agent treatment process, ensuring that the sizing agent penetration reaches an ideal effect and is evenly distributed. When the value of the comprehensive index of the sizing agent penetration quality exceeds a certain threshold, it indicates that the sizing agent penetration process meets the expectations, otherwise, the process parameters need to be further optimized. According to the comprehensive index of the sizing agent penetration quality, a combination of sizing agent deep penetration process parameters for the chenille yarn steaming process is generated. By analyzing the process parameters under different CQI values, a set of optimal sizing agent deep penetration process parameters are obtained, including factors such as steam temperature, pressure, humidity, etc. By optimizing and adjusting these parameters, the penetration effect of the sizing agent is improved, ensuring that the quality of the final product meets the requirements.

[0031] In this embodiment, the penetration data under different process parameters are collected, and these data are correlated with the comprehensive index of the penetration quality of the sizing agent. Through data analysis methods, a multi-dimensional relationship model including steam pressure, temperature, humidity, and treatment time is established to reveal how each parameter affects different characteristics in the sizing agent penetration process. Parameter sensitivity analysis is performed on the influence relationship model to determine the degree of influence of each process parameter on the comprehensive index of the penetration quality of the sizing agent. By calculating partial derivatives, the priority of different parameters is determined by analyzing the influence of the change of each process parameter on the comprehensive index of the penetration quality. Through gradient calculation, the optimal amplitude of adjustment for each parameter is determined to ensure that, on the premise of meeting the requirements of the penetration effect, the adjustment amplitude can effectively improve the penetration effect of the sizing agent without causing unnecessary waste of resources. A fuzzy control rule base is constructed based on the expert knowledge and historical data of the chenille yarn steaming process, converting expert experience and historical production data into operable rules, and realizing the intelligent adjustment of steam parameters through fuzzy control methods. The fuzzy control rule base consists of a series of IF-THEN rules. The input of the rules is the measured value and trend of the current penetration state, such as indicators like the comprehensive index of the penetration quality of the sizing agent, penetration depth, and uniformity. The output of the rules is the adjustment suggestions for steam parameters (such as temperature, humidity, pressure, and treatment time). With the support of the fuzzy control rule base, the automatic execution of the steam parameter adjustment strategy is realized. This strategy dynamically adjusts steam parameters by real-time monitoring the penetration state in the production process and combining the current comprehensive index of the penetration quality. When the penetration state of the sizing agent does not meet the expectation, parameters such as steam pressure, temperature, humidity, and treatment time are adjusted according to the information in the rule base to promote the deep penetration of the sizing agent and ensure the final penetration effect. At the same time, through learning from historical data, the fuzzy control system can continuously optimize the rule base and improve the accuracy and efficiency of the control strategy. Through the above steps, based on the comprehensive index of the penetration quality of the sizing agent and parameter sensitivity analysis, a combination of process parameters for the deep penetration of the sizing agent for the chenille yarn steaming process is generated.

[0032] In the embodiments of the present application, through the combination of multi-wavelength near-infrared spectroscopy and infrared thermal imaging technology, the accurate distinction and monitoring of the surface adhesion amount and internal penetration amount of the sizing agent are realized, overcoming the technical bottleneck that the traditional method cannot distinguish the distribution levels of the sizing agent. The multi-layer penetration feature vector of the sizing agent constructed by the spectral-temperature transformation method realizes the comprehensive characterization of the distribution states of the sizing agent in the surface layer, shallow layer and deep layer of the fabric. The sizing agent penetration time series recognition network analyzes the continuously collected feature data, can capture the dynamic changes in the sizing agent penetration process in real time, and effectively distinguish different penetration states. The dual-task evaluation model of penetration and uniformity realizes the parallel evaluation and comprehensive consideration of the penetration depth and uniformity through the shared coding layer and feature interaction mechanism. The comprehensive sizing agent penetration quality index generated provides a quantitative basis for the optimization of process parameters. The parameter mapping relationship and fuzzy control rule base established based on this comprehensive index can automatically generate the optimal process parameter combination according to the characteristics of different chenille fabrics, realizing the closed-loop control from monitoring data to process parameters, improving the accuracy and stability of sizing agent treatment. At the same time, by precisely controlling the sizing agent penetration process, over-steaming of yarn and sizing agent waste are avoided.

[0033] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0034] Install and calibrate the near-infrared spectroscopy sensor array to obtain a multi-point near-infrared detection system, and position and arrange the infrared thermal imaging device to obtain a temperature distribution monitoring system;

[0035] Collect the original spectral data of the chenille steamed yarn fabric during the steaming process through the multi-point near-infrared detection system, and perform baseline correction and scattering correction on the original spectral data to obtain the near-infrared spectral data reflecting the molecular structure characteristics of the sizing agent;

[0036] Collect the original thermal imaging data of the chenille steamed yarn fabric during the steaming process through the temperature distribution monitoring system, and perform temperature gradient calculation on the original thermal imaging data to obtain the infrared thermal imaging data characterizing the penetration depth of the sizing agent;

[0037] Perform data fusion on the near-infrared spectral data and the infrared thermal imaging data to obtain the sizing agent double-layer distribution monitoring data characterizing the surface adhesion amount and internal penetration amount of the sizing agent.

[0038] Specifically, each component of the system is designed and configured. When installing the near-infrared spectral sensor array, the distribution method, installation location, and distance from the fabric are considered. The sensor array includes at least 8 near-infrared detectors with a wavelength range of 900 - 1700 nm and a resolution of not less than 2 nm, and these detectors are annularly distributed and installed at the inlet and outlet of the yarn steaming machine. To ensure accurate collection, the distance between the sensor and the chenille fabric is maintained between 5 ± 0.5 cm. During installation, ensure the consistency of the installation location and spacing of each detector to ensure full coverage of all areas of the fabric during the yarn steaming process. At the same time, wavelength calibration and sensitivity calibration are performed on the spectral sensor to ensure that the collected data has high accuracy and high consistency, and to avoid affecting subsequent analysis results due to calibration errors. Position and arrange the infrared thermal imaging device. The infrared thermal imaging device can monitor the temperature distribution on the fabric surface in real time to help analyze the penetration depth of the sizing agent. To ensure the accuracy of temperature monitoring, an infrared thermal imaging device with high thermal resolution is selected, such as an infrared thermal camera with a thermal resolution of not less than 0.05 °C. The acquisition frequency of this device is set to 10 frames per second to quickly capture the subtle temperature changes during the yarn steaming process. The device is installed at the center position of the near-infrared sensor array to ensure synchronous capture of comprehensive temperature distribution data. This arrangement ensures the coordinated operation of the two systems (near-infrared spectral sensor array and infrared thermal imaging device). The original spectral data of the chenille yarn-steamed fabric during the yarn steaming process is collected through a multi-point near-infrared detection system, reflecting the distribution of the sizing agent on the fabric surface and inside. Since the spectral data is affected by background noise and light scattering effects, data preprocessing is performed on the original spectral data, including baseline correction and scattering correction. Baseline correction can eliminate the background signal in the spectral data, making the data more accurately reflect the characteristics of the sizing agent molecular structure. Scattering correction is used to eliminate the influence of light scattering effects on the spectral signal, so that the collected data is only related to the sizing agent molecular structure. The corrected spectral data can provide the adhesion of the sizing agent on the fabric surface and its penetration at different depths, helping to more accurately analyze the penetration process of the sizing agent. At the same time, the original thermal imaging data collected by the temperature distribution monitoring system through the infrared thermal imaging device is used to calculate the temperature gradient. The temperature gradient is a key indicator for analyzing the penetration depth of the sizing agent because the penetration of the sizing agent will affect the temperature distribution in different depth regions of the fabric. By calculating the temperature gradient in the thermal imaging data, a curve describing the temperature change rate is obtained, and thus the penetration depth of the sizing agent is inferred based on the degree of temperature change. For example, the temperature change on the surface layer is large, while the temperature change in the deep layer is small, reflecting the penetration difference of the sizing agent between the surface layer and the deep layer. The calculation result of the temperature gradient can help analyze the penetration state of the sizing agent. Data fusion is performed on the near-infrared spectral data and the infrared thermal imaging data to comprehensively characterize the distribution of the sizing agent in the chenille yarn-steamed fabric, specifically including the adhesion amount of the sizing agent on the fabric surface and the penetration amount at different depth levels.By integrating the characteristic peaks of the sizing agent in the spectral data and the temperature gradient information in the temperature data, the accurate quantification of the sizing agent distribution is achieved. For example, in the near-infrared spectral data, the sizing agent usually exhibits characteristic absorption peaks at 1200 nm, 1400 nm, and 1600 nm, while the temperature data can provide the variation information of the surface temperature and the deep temperature. By combining these data, a more comprehensive monitoring map of the sizing agent distribution is obtained, which not only includes the adhesion amount on the surface layer but also reveals the penetration degree of the sizing agent in the shallow layer and the deep layer. For example, if the sizing agent absorption peak intensity at a certain wavelength is large in the spectral data, and the infrared thermal imaging data shows that the temperature in this area is relatively high, it is inferred that the penetration of the sizing agent in this area is shallow and mainly concentrated in the surface layer. On the contrary, if the spectral data indicates that the absorption peak in the deep area is stronger, and the temperature data shows that the temperature change in this area is smaller, it is speculated that the sizing agent has successfully penetrated to a deeper level.

[0039] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0040] Perform standard normal variate transformation on the near-infrared spectral data in the sizing agent bilayer distribution monitoring data to obtain characteristic spectral data, and perform characteristic peak identification and analysis on the characteristic spectral data to obtain characteristic peak intensity change data;

[0041] Construct a temperature gradient model characterizing the heat conduction characteristics based on the infrared thermal imaging data in the sizing agent bilayer distribution monitoring data;

[0042] Perform hierarchical division processing on the chenille yarn steaming fabric based on the temperature gradient model to obtain a hierarchical data structure of three regions: the surface layer, the shallow layer, and the deep layer;

[0043] Calculate the sizing agent distribution index for each layer according to the characteristic peak intensity change data and the hierarchical data structure to obtain target characteristic data including the sizing agent distribution index, the temperature change rate, and the spectral standard deviation, and perform principal component analysis on the target characteristic data to obtain the sizing agent multi-layer penetration characteristic vector.

[0044] Specifically, standard normal variate transformation is performed on the near-infrared spectral data in the monitoring data of the double-layer distribution of the oil agent to eliminate the systematic errors caused by sample and instrument factors, so that the spectral data has the same mean and variance among different sampling points, highlighting the oil agent characteristics in the spectral data. In the spectral data after standard normal variate transformation, characteristic absorption peaks related to the molecular structure of the oil agent are identified, especially the characteristic peaks at 1200 nm, 1400 nm, and 1600 nm. These peak values reflect the characteristic absorption of the oil agent in the near-infrared spectrum. By analyzing the intensity changes of these characteristic peaks, information related to the oil agent distribution is obtained, such as the penetration depth and surface adhesion amount of the oil agent at different levels. At the same time, a temperature gradient model is constructed using the original thermal imaging data collected by the infrared thermal imaging device. Since the penetration of the oil agent will affect the heat conduction characteristics of the fabric, the penetration state of the oil agent at different depths will cause different temperature changes on the fabric surface and in different depth regions. By calculating the temperature gradient in the thermal imaging data, a temperature gradient model characterizing the heat conduction characteristics is obtained. This model helps to determine the penetration of the oil agent at each depth level. The temperature change on the surface layer is usually more drastic, while the temperature change in the deep layer region is relatively gentle. The temperature gradient model helps the system accurately evaluate the penetration depth of the oil agent by analyzing these temperature changes. Based on the temperature gradient model, the chenille yarn steaming fabric is divided into three regions: the surface layer, the shallow layer, and the deep layer. The surface layer is located at a depth of 0 - 0.5 mm, the shallow layer is located at 0.5 - 1.5 mm, and the deep layer is located in the region greater than 1.5 mm. Through this hierarchical division, the oil agent distribution in each region is calculated separately, reflecting the penetration status of the oil agent at different depth layers. According to the characteristic peak intensity change data and the hierarchical data structure, the oil agent distribution index (ODI) of each layer is calculated. The oil agent distribution index ODI_s of the surface layer is calculated through near-infrared reflection spectroscopy, the oil agent distribution index ODI_m of the shallow layer is calculated by combining diffuse reflection spectroscopy and temperature gradient data, and the oil agent distribution index ODI_d of the deep layer is calculated by combining the temperature decay curve and the second derivative of the spectrum. These oil agent distribution indexes can accurately reflect the penetration degree of the oil agent on the fabric surface and inside. Data such as the oil agent distribution index, temperature change rate, and spectral standard deviation are used as target characteristic data for principal component analysis. Principal component analysis is a dimensionality reduction technique that transforms multi-dimensional characteristic data into fewer principal components while retaining the main information in the data. Through principal component analysis, the most representative characteristics are extracted from the complex multi-dimensional monitoring data to characterize the distribution state of the oil agent in the chenille fabric. Through principal component analysis, a characteristic vector containing multiple principal components is obtained, reflecting the overall situation of the oil agent penetration and the penetration depth of different levels. For example, through the fusion analysis of the characteristic peak intensity change data and the temperature gradient data, the distribution characteristics of the oil agent in the surface layer, shallow layer, and deep layer are identified.If the characteristic peak intensity of the surface layer is large and the temperature change is relatively drastic, it indicates that the sizing agent is mainly concentrated in the surface layer; if the characteristic peak intensities of the shallow layer and the deep layer are small and the temperature change is small, it indicates that the penetration of the sizing agent is shallow and mainly stays on the surface. In this way, the penetration state of the sizing agent is dynamically monitored, and the process is optimized based on these characteristic data to ensure the uniformity and depth of the penetration of the sizing agent.

[0045] In a specific embodiment, the process of performing hierarchical division processing on the chenille yarn steaming fabric based on the temperature gradient model to obtain the hierarchical data structure of the three regions of the surface layer, the shallow layer, and the deep layer may specifically include the following steps:

[0046] Calculate the heat conduction rate for each spatial position point of the temperature gradient model to obtain the temperature decay curve function characterizing the heat transfer characteristics in the chenille yarn steaming fabric;

[0047] Determine the temperature decay rate mutation point according to the temperature decay curve function to obtain the depth critical value for distinguishing different physical structure layers;

[0048] Perform adaptive matching between the depth critical value and the preset depth demarcation point to obtain the correction coefficient for the current chenille yarn steaming fabric structure;

[0049] Perform three-dimensional reconstruction on the temperature gradient model based on the correction coefficient to obtain a three-dimensional temperature distribution model including spatial coordinates and temperature values;

[0050] Calculate the inter-layer temperature gradient of the three-dimensional temperature distribution model to obtain the temperature change rate matrix characterizing the penetration state of the sizing agent in each layer;

[0051] Construct the characteristic mapping relationship of the three-layer structure according to the temperature change rate matrix and the spectral second derivative data to obtain the hierarchical data structure of the three regions of the surface layer, the shallow layer, and the deep layer.

[0052] Specifically, the temperature gradient model reflects the penetration of the sizing agent in the chenille yarn-steaming fabric by measuring the temperature changes in different depth regions. By calculating the temperature changes at each spatial position point, the heat conduction characteristics of the fabric are obtained. This process includes generating a temperature decay curve based on the temperature changes and the heat conduction characteristics of the fabric, which describes how heat is transferred from the steam treatment process to the surface and interior of the fabric. The shape of the temperature decay curve depends on the penetration degree of the sizing agent in the fabric and the heat conduction efficiency. Therefore, the temperature decay curve can effectively reflect the penetration depth of the sizing agent. The critical depth value for distinguishing different physical structure layers is obtained by determining the mutation point of the temperature decay rate according to the temperature decay curve function. The temperature decay curve shows different slopes in different depth regions. Especially in the surface layer, shallow layer, and deep layer regions of the fabric, there are significant differences in the rate of temperature change. The mutation point of the temperature decay rate appears at the physical boundary of the sizing agent penetration, such as between the surface layer and the shallow layer, and between the shallow layer and the deep layer. By analyzing the slope change of the temperature decay curve, these mutation points are identified and determined as the critical depth values for distinguishing different physical structure layers. The critical depth value is adaptively matched with the preset depth demarcation point to obtain a correction coefficient for the current chenille yarn-steaming fabric structure. The preset demarcation point is a standard value obtained based on the general characteristics of chenille fabrics or historical data. Due to the differences in different batches of fabrics and sizing agent treatments, these preset demarcation points need to be adjusted to better adapt to the actual situation of the current fabric. Through the adaptive matching algorithm, these demarcation points are adjusted according to the temperature decay characteristics of the current fabric to obtain a correction coefficient for this fabric. Based on the correction coefficient, a three-dimensional reconstruction of the temperature gradient model is performed to obtain a three-dimensional temperature distribution model containing spatial coordinates and temperature values. The temperature gradient model itself is a two-dimensional model that contains temperature change data in the depth direction. By applying the correction coefficient to the temperature data, the model is extended to three-dimensional spatial coordinates, and each spatial point has a corresponding temperature value, so that the temperature distribution is not limited to the depth direction but can be comprehensively displayed within the spatial range of the fabric. The interlayer temperature gradient of the three-dimensional temperature distribution model is calculated to obtain a temperature change rate matrix representing the penetration state of the sizing agent in each layer. The temperature gradient calculation is carried out based on the temperature difference between different depth layers in the model. By calculating the temperature change rate within each depth layer, a temperature change rate matrix is obtained, which further represents the penetration state of the sizing agent within each layer. The temperature change rate in the surface layer is relatively large, while that in the deep layer is relatively small, which reflects the penetration depth and uniformity of the sizing agent. The temperature change rate matrix helps the system identify the regions with uneven sizing agent penetration and reveals the distribution characteristics of the sizing agent in different regions. A characteristic mapping relationship of a three-layer structure is constructed based on the temperature change rate matrix and the spectral second derivative data to obtain a hierarchical data structure of the surface layer, shallow layer, and deep layer regions.Combine the temperature change rate matrix with the second derivative data reflecting the distribution characteristics of the oil agent in the near-infrared spectral data to establish a characteristic mapping relationship with a three-layer structure. This relationship clearly links the penetration state and distribution characteristics of the oil agent with the physical levels of the fabric, obtaining a hierarchical data structure for the surface layer, shallow layer, and deep layer. Through this step, the distribution of the oil agent in different depth regions is described.

[0053] In a specific embodiment, the process of performing step S103 may specifically include the following steps:

[0054] Input the multi-layer penetration feature vector of the oil agent into the input layer of the oil agent penetration time series recognition network to obtain a feature sequence data stream. The oil agent penetration time series recognition network includes an input layer, a first LSTM hidden layer, a second LSTM hidden layer, and an output layer;

[0055] Perform forward information extraction and backward information extraction on the feature sequence data stream in the first LSTM hidden layer to obtain an intermediate feature vector containing the state of the gating unit;

[0056] Transfer the intermediate feature vector to the second LSTM hidden layer for advanced feature extraction to obtain the time series state features characterizing the dynamic process of oil agent penetration, and map the time series state features to a three-dimensional state space;

[0057] Calculate the oil agent penetration state vector based on the output value of the three-dimensional state space to obtain the probability distributions of the surface attachment state, penetration progress state, and penetration completion state. When the probability value of the penetration completion state is greater than the preset target value, it is determined as the oil agent penetration completion state;

[0058] In the output layer, calculate the effective penetration amount of the oil agent according to the feature vector corresponding to the oil agent penetration completion state to obtain a quantitative penetration depth index characterizing the distribution ratio of the oil agent in each layer of the chenille yarn steaming fabric.

[0059] Specifically, the oil agent multi-layer penetration feature vector is input into the input layer of the oil agent penetration time series recognition network to obtain the feature sequence data stream. The oil agent multi-layer penetration feature vector contains information such as the oil agent distribution index, temperature change rate, and spectral standard deviation extracted from near-infrared spectroscopy data and infrared thermal imaging data. These feature vectors represent the distribution of the oil agent on the surface and at different depth levels of the chenille yarn steaming fabric. The structure of the oil agent penetration time series recognition network consists of multiple layers, including an input layer, two LSTM (Long Short-Term Memory) hidden layers, and an output layer. The input layer receives the processed feature vector sequence data, which represents the penetration state of the oil agent at multiple time points. Through these time series data, the LSTM network captures the time series changes and dynamic processes of the oil agent penetration. When the first LSTM hidden layer processes the feature sequence data stream, it adopts the methods of forward information extraction and backward information extraction. In the forward information extraction process, the LSTM calculates a new hidden state based on the current input feature and the state information of the previous moment, while the backward information extraction is to optimize and adjust the state by using the signals of subsequent time steps during backpropagation. The bidirectional information extraction method enables the LSTM to capture long-term and short-term dependencies in the time series and extract intermediate feature vectors related to the dynamic process of the oil agent penetration. The intermediate feature vectors contain the key information in the oil agent penetration process at each time point, reflect the penetration state at the current moment, and retain the influence of historical time points, thus improving the accuracy of the penetration process recognition. The intermediate feature vectors are passed to the second LSTM hidden layer for the extraction of high-level features. In the second LSTM, the network processes the intermediate feature vectors to extract more high-level time series features. The output of the second LSTM can effectively characterize the time change trend and depth information in the dynamic process of the oil agent penetration. Through the feature extraction of this layer, the network understands the whole picture of the oil agent penetration, including the entire process from surface attachment to deep penetration. The time series state features are mapped to a three-dimensional state space to reflect the penetration state of the oil agent on the fabric surface and at different depth levels. In the three-dimensional space, the output value of each point corresponds to different stages of the oil agent penetration. These stages include different states of surface attachment, penetration progress, and penetration completion. Based on the output values in the three-dimensional state space, the network calculates the penetration state vector of the oil agent. The three components of this vector respectively represent the probability distributions of the surface attachment state, penetration progress state, and penetration completion state. For example, during the penetration progress state, the penetration amount of the oil agent will gradually increase, while in the penetration completion state, the oil agent has reached the predetermined penetration depth. When the probability value of the penetration completion state exceeds the set target value, such as set to 0.7, the network determines that the oil agent penetration has been completed. Through this determination, the penetration state of the oil agent is monitored in real time and feedback is provided to the operator in a timely manner. In the output layer, based on the feature vector corresponding to the penetration completion state, the effective penetration amount of the oil agent is calculated. The effective penetration amount reflects the actual penetration degree of the oil agent in the chenille yarn steaming fabric, especially the penetration amount in the deep layer.For example, when the probability value of the penetration completion state is high enough, the effective penetration amount of the oil agent is calculated through a preset calibration curve based on the eigenvector corresponding to this state. At this time, the weight coefficients in the calibration curve are optimized through experiments. For example, the weight coefficients of the oil agent distribution in the surface layer, the shallow layer, and the deep layer are 0.6, 0.3, and 0.1 respectively. The effective penetration amount calculated through these coefficients accurately reflects the penetration ratio of the oil agent in each layer.

[0060] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0061] Perform multi-point sampling on the chenille yarn steaming fabric based on the quantitative penetration depth index to obtain the penetration depth distribution data in the width direction of the fabric;

[0062] Input the quantitative penetration depth index into the shared encoding layer of the dual-task neural network to obtain an intermediate feature representation;

[0063] Enhance the intermediate feature representation through the feature interaction module using the attention mechanism to obtain an enhanced feature with information interaction between tasks;

[0064] Input the enhanced feature into the penetration state decoding layer and the uniformity decoding layer respectively for dual decoding to obtain a penetration state evaluation value and a uniformity evaluation value;

[0065] Calculate the overall effective penetration amount of the fabric according to the penetration state evaluation value, and calculate the coefficient of variation of the penetration amount in the width direction of the fabric according to the uniformity evaluation value to obtain the comprehensive index of the oil agent penetration quality;

[0066] Generate a combination of deep penetration process parameters of the oil agent for the chenille yarn steaming process according to the comprehensive index of the oil agent penetration quality.

[0067] Specifically, through the spectral monitoring and temperature distribution monitoring data in the system, the penetration depth data of chenille yarn steaming fabric at different positions are obtained. The quantitative penetration depth index can describe the penetration degree of the sizing agent in the fabric, including the penetration levels at different depths. These data are obtained by multi-point sampling in the width direction of the fabric, and the sampling points are evenly distributed throughout the entire width range of the fabric. Through multi-point sampling, the distribution of the penetration depth of the sizing agent in the width direction of the fabric is obtained, reflecting the penetration of the sizing agent in different regions. The quantitative penetration depth index is input into the shared encoding layer of the dual-task neural network. The role of the shared encoding layer is to fuse these penetration depth data and other relevant features (such as temperature, sizing agent distribution index, etc.) to obtain a unified intermediate feature representation, including the penetration state of the sizing agent in the fabric and other characteristics of the sizing agent distribution. Through the integration of these data, the neural network can learn multi-dimensional penetration information. The intermediate features are processed by the feature interaction module in the shared encoding layer. The feature interaction module enhances the information interaction between tasks through the attention mechanism. In the dual-task learning framework, the penetration state recognition task and the uniformity evaluation task share the features in the shared encoding layer. Therefore, the feature interaction module automatically learns which features are more important in these two tasks through the attention mechanism. In this way, the attention mechanism assigns different weights to different features, enhances the features related to the sizing agent penetration process, and weakens the features unrelated to uniformity. The introduction of the attention mechanism enables the model to adaptively strengthen the connection between different tasks and improve the performance of the model in multi-task learning. The enhanced features are respectively input into the penetration state decoding layer and the uniformity decoding layer. The penetration state decoding layer focuses on the classification and recognition of the penetration state of the sizing agent, including states such as surface attachment, penetration in progress, and penetration completion. In this layer, the network outputs the penetration state evaluation value according to the input enhanced features. The uniformity decoding layer focuses on calculating the uniformity of the sizing agent penetration. By analyzing the penetration depth of each sampling point in the width direction of the fabric, it outputs the uniformity evaluation value. According to the penetration state evaluation value, the overall effective penetration amount of the fabric is calculated. The effective penetration amount is a key index, reflecting the actual penetration degree of the sizing agent in the fabric, especially the penetration amount in the deep region. Through the penetration state evaluation value, it is determined whether the sizing agent has penetrated to the predetermined depth of the fabric, and the effective penetration amount is calculated. And according to the uniformity evaluation value, the coefficient of variation of the penetration amount in the width direction of the fabric is calculated. This coefficient of variation describes the penetration uniformity of the sizing agent at different positions of the fabric. If the coefficient of variation is small, it indicates that the sizing agent penetrates evenly; if the coefficient of variation is large, it means that the penetration in some regions is too shallow or too deep, thus affecting the overall effect of the sizing agent treatment. Through the calculated effective penetration amount and coefficient of variation, the comprehensive quality index of sizing agent penetration (CQI) is obtained. CQI is a comprehensive evaluation index for the penetration quality of the sizing agent, combining the effective penetration amount and penetration uniformity.The calculation of the CQI takes into account the influence of the effective penetration amount on the distribution of the deep-layer oil agent in the fabric and the influence of the coefficient of variation on the penetration uniformity. Through this index, the penetration quality of the oil agent can be quickly evaluated. For example, if the CQI value is greater than 3.0, it indicates that the penetration effect of the oil agent is ideal and the uniformity is good; if the CQI value is between 2.5 and 3.0, it means that the penetration effect is good, but there are slight uniformity problems in some areas; if the CQI value is lower than 2.5, it means that there are obvious problems with the penetration of the oil agent and the process parameters need to be adjusted. According to the comprehensive index of the oil agent penetration quality (CQI), a combination of deep-layer penetration process parameters for the chenille steaming process is generated. By matching different CQI values with the corresponding process parameters, the best combination of oil agent penetration process parameters is found. These parameters include the temperature, humidity, pressure of the steam, etc. By reasonably adjusting these process parameters, it is ensured that the oil agent penetrates evenly and to an appropriate depth, ultimately optimizing the quality of the fabric.

[0068] In a specific embodiment, the process of generating a combination of deep-layer penetration process parameters for the chenille steaming process according to the comprehensive index of the oil agent penetration quality may specifically include the following steps:

[0069] Based on the comprehensive index of the oil agent penetration quality, a mapping relationship between the steam parameters and the oil agent penetration performance is established, and an influence relationship model of the four process parameters of steam pressure, temperature, humidity, and treatment time on the penetration effect is obtained;

[0070] Perform a parameter sensitivity analysis on the influence relationship model to obtain the ranking of the influence degrees of each parameter on the comprehensive index of the oil agent penetration quality, and determine the priority order and the best adjustment range of parameter adjustment through gradient calculation;

[0071] Construct a fuzzy control rule base according to the expert knowledge and historical data of the chenille steaming process, obtain the steam parameter adjustment strategy, and perform parameter adjustment calculation based on the fuzzy control rule base to obtain a combination of deep-layer penetration process parameters for the oil agent.

[0072] Specifically, a multi-parameter mathematical model is established through experimental data and production data to describe the influence of four key process parameters, namely steam pressure, temperature, humidity, and treatment time, on the penetration effect of the finishing agent. By collecting data on chenille fabrics treated under different steam conditions, especially their responses during the penetration process of the finishing agent, the relationship between these process parameters and the penetration quality of the finishing agent is found. Based on these data, a model reflecting the influence relationship between steam parameters and the penetration performance of the finishing agent is established. This model provides a systematic way to analyze and predict the changing trends of the penetration effect of the finishing agent under different process conditions. For example, under different temperature and humidity conditions, how will the penetration depth and uniformity of the finishing agent change, thereby affecting the quality of the final fabric. Perform a parameter sensitivity analysis on the established influence relationship model to evaluate the influence degree of different process parameters on the comprehensive index of the penetration quality of the finishing agent, so as to determine the contribution of each parameter to the penetration effect. Through the parameter sensitivity analysis, quantify the influence of each process parameter on the penetration quality and rank them. Through gradient calculation, determine the priority order of adjusting each process parameter, clarify which parameters should be adjusted first, and the gradient calculation can also help determine the optimal adjustment amplitude for each parameter to ensure that resources are not wasted or over-treatment occurs while improving the penetration effect. Through this step, achieve fine control of the process parameters to ensure the maximization of the penetration effect of the finishing agent. According to the expert knowledge and historical data of the chenille steaming process, construct a fuzzy control rule base. The construction of the fuzzy control rule base relies on empirical knowledge and a large amount of production data. Through this rule base, reasonable steam parameter adjustment strategies are designed for different production scenarios. The fuzzy control rules consist of "if-then" type rules. These rules better adapt to the complex situations and changes in actual production by combining expert knowledge and historical data. For example, when the penetration effect of the finishing agent on some batches of chenille fabrics is not ideal during the production process, corresponding adjustment strategies are proposed according to the existing rule base, such as increasing the steam temperature or extending the treatment time. Through these fuzzy control rules, make adaptive adjustments under different production stages and conditions to ensure that the penetration process of the finishing agent is stable and meets expectations. Based on the fuzzy control rule base, perform parameter adjustment calculations to obtain the parameter combination for deep penetration of the finishing agent. These adjustment strategies consider various factors affecting the penetration of the finishing agent, such as temperature, humidity, pressure, and time. By optimizing these parameter combinations, the best penetration effect is achieved. The parameter combination specifically guides the operation of the production line to ensure that the quality of the penetration of the finishing agent remains stable during the production of different batches. For example, assume that the comprehensive index of the penetration quality of the finishing agent (CQI) is low. According to the analysis of the influence relationship model and the sensitivity analysis, it is found that temperature has a greater impact on the penetration depth, while humidity has a greater impact on the uniformity. Then the system adjusts the temperature, increases the steam temperature to improve the penetration depth, and fine-tunes the humidity through the fuzzy control rule base to ensure the uniformity of the penetration of the finishing agent. After adjustment, both the penetration depth and uniformity are improved, and the CQI value rebounds to the ideal level.

[0073] The intelligent monitoring method for the chemical fiber oil agent of chenille yarn steaming fabric in the above embodiments of the present application has been described. Next, the intelligent monitoring system for the chemical fiber oil agent of chenille yarn steaming fabric in the embodiments of the present application will be described. Please refer to Figure 2 One embodiment of the intelligent monitoring system for the chemical fiber oil agent of chenille yarn steaming fabric in the embodiments of the present application includes:

[0074] An acquisition module 201, configured to perform multi-wavelength near-infrared spectroscopy and infrared thermal imaging acquisition on the chenille yarn steaming fabric during the yarn steaming process to obtain oil agent double-layer distribution monitoring data;

[0075] A transformation module 202, configured to perform a spectrum-temperature transformation on the oil agent double-layer distribution monitoring data to obtain an oil agent multi-layer penetration feature vector;

[0076] A state analysis module 203, configured to input the oil agent multi-layer penetration feature vector into an oil agent penetration time series recognition network for state analysis to obtain a quantitative penetration depth index;

[0077] A generation module 204, configured to perform a dual-task evaluation of penetration and uniformity based on the quantitative penetration depth index to generate an oil agent deep penetration process parameter combination for the chenille yarn steaming process.

[0078] Through the collaborative cooperation of the above-mentioned various components, through the combination of multi-wavelength near-infrared spectroscopy and infrared thermal imaging technologies, accurate discrimination monitoring of the surface adhesion amount and internal penetration amount of the oil agent is realized, overcoming the technical bottleneck that the traditional method cannot distinguish the oil agent distribution levels. The oil agent multi-layer penetration feature vector constructed by the spectrum-temperature transformation method realizes a comprehensive characterization of the oil agent distribution states in the three regions of the fabric surface layer, shallow layer, and deep layer. The oil agent penetration time series recognition network analyzes the continuously collected feature data, can capture the dynamic changes of the oil agent penetration process in real time, and effectively distinguish different penetration states. The dual-task evaluation model of penetration and uniformity realizes the parallel evaluation and comprehensive consideration of penetration depth and uniformity through a shared encoding layer and a feature interaction mechanism. The generated comprehensive oil agent penetration quality index provides a quantitative basis for process parameter optimization. The parameter mapping relationship and fuzzy control rule base established based on this comprehensive index can automatically generate the optimal process parameter combination according to the characteristics of different chenille fabrics, realizing the closed-loop control from monitoring data to process parameters, improving the accuracy and stability of oil agent treatment, and at the same time avoiding over-steaming of yarn and waste of oil agent by precisely controlling the oil agent penetration process.

[0079] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the above-described system, system, and units can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0080] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing an intelligent monitoring device for chenille steaming and sizing fabric chemical fiber oil agent (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs.

[0081] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of various embodiments of the present invention.

Claims

1. An intelligent monitoring method for chemical fiber oil agents of chenille steamed yarn fabric, characterized in that, The method includes: Performing multi-wavelength near-infrared spectroscopy and infrared thermal imaging acquisition on the chenille yarn steaming fabric during the yarn steaming process to obtain oil agent bilayer distribution monitoring data; Performing a spectrum-temperature transformation on the oil agent bilayer distribution monitoring data to obtain an oil agent multi-layer penetration feature vector; Inputting the oil agent multi-layer penetration feature vector into an oil agent penetration time series recognition network for state analysis to obtain a quantitative penetration depth index; Performing a dual-task evaluation of penetration and uniformity based on the quantitative penetration depth index to generate an oil agent deep penetration process parameter combination for the chenille yarn steaming process.

2. The intelligent monitoring method for the chemical fiber finishing agent of chenille yarn steaming fabric according to claim 1, wherein, The performing multi-wavelength near-infrared spectroscopy and infrared thermal imaging acquisition on the chenille yarn steaming fabric during the yarn steaming process to obtain oil agent bilayer distribution monitoring data includes: Installing and calibrating a near-infrared spectroscopy sensor array to obtain a multi-point near-infrared detection system, and positioning and arranging an infrared thermal imaging device to obtain a temperature distribution monitoring system; Collecting original spectral data of the chenille yarn steaming fabric during the yarn steaming process through the multi-point near-infrared detection system, and performing baseline correction and scattering correction on the original spectral data to obtain near-infrared spectral data reflecting the molecular structure characteristics of the oil agent; Collecting original thermal imaging data of the chenille yarn steaming fabric during the yarn steaming process through the temperature distribution monitoring system, and performing a temperature gradient calculation on the original thermal imaging data to obtain infrared thermal imaging data characterizing the oil agent penetration depth; Performing data fusion on the near-infrared spectral data and the infrared thermal imaging data to obtain oil agent bilayer distribution monitoring data characterizing the surface adhesion amount and internal penetration amount of the oil agent.

3. The intelligent monitoring method for the chemical fiber oil agent of chenille yarn steaming fabric according to claim 1, characterized in that The performing a spectrum-temperature transformation on the oil agent bilayer distribution monitoring data to obtain an oil agent multi-layer penetration feature vector includes: Performing a standard normal variate transformation on the near-infrared spectral data in the oil agent bilayer distribution monitoring data to obtain characteristic spectral data, and performing characteristic peak identification analysis on the characteristic spectral data to obtain characteristic peak intensity change data; Constructing a temperature gradient model characterizing the heat conduction characteristics according to the infrared thermal imaging data in the oil agent bilayer distribution monitoring data; Performing a hierarchical division process on the chenille yarn steaming fabric based on the temperature gradient model to obtain a hierarchical data structure of three regions: the surface layer, the shallow layer, and the deep layer; Calculating the oil agent distribution index of each layer according to the characteristic peak intensity change data and the hierarchical data structure to obtain target characteristic data including the oil agent distribution index, the temperature change rate, and the spectral standard deviation, and performing principal component analysis on the target characteristic data to obtain an oil agent multi-layer penetration feature vector.

4. The intelligent monitoring method for the chemical fiber oil agent of chenille steamed yarn fabric according to claim 3, characterized in that, The performing a hierarchical division process on the chenille yarn steaming fabric based on the temperature gradient model to obtain a hierarchical data structure of three regions: the surface layer, the shallow layer, and the deep layer includes: Performing a heat conduction rate calculation on each spatial position point of the temperature gradient model to obtain a temperature decay curve function characterizing the heat transfer characteristics in the chenille yarn steaming fabric; Determining a temperature decay rate mutation point according to the temperature decay curve function to obtain a depth critical value for distinguishing different physical structure layers; Adaptive matching is performed between the depth critical value and a preset depth demarcation point to obtain a correction coefficient for the current chenille yarn steaming fabric structure; Based on the correction coefficient, three-dimensional reconstruction is performed on the temperature gradient model to obtain a three-dimensional temperature distribution model including spatial coordinates and temperature values; Interlayer temperature gradient calculation is performed on the three-dimensional temperature distribution model to obtain a temperature change rate matrix characterizing the oil agent penetration state of each layer; According to the temperature change rate matrix and spectral second derivative data, a characteristic mapping relationship of a three-layer structure is constructed to obtain a hierarchical data structure of three regions: the surface layer, the shallow layer, and the deep layer.

5. The intelligent monitoring method for the chemical fiber oil agent of chenille steaming yarn fabric according to claim 1, characterized in that, Inputting the oil agent multi-layer penetration feature vector into the oil agent penetration time series recognition network for state analysis to obtain a quantitative penetration depth index, including: Inputting the oil agent multi-layer penetration feature vector into the input layer of the oil agent penetration time series recognition network to obtain a feature sequence data stream. The oil agent penetration time series recognition network includes an input layer, a first LSTM hidden layer, a second LSTM hidden layer, and an output layer; Forward information extraction and backward information extraction are performed on the feature sequence data stream in the first LSTM hidden layer to obtain an intermediate feature vector including the state of the gating unit; The intermediate feature vector is passed to the second LSTM hidden layer for advanced feature extraction to obtain a time series state feature characterizing the dynamic process of oil agent penetration, and the time series state feature is mapped to a three-dimensional state space; Based on the output value of the three-dimensional state space, an oil agent penetration state vector is calculated to obtain the probability distributions of the surface attachment state, the penetration progress state, and the penetration completion state. When the probability value of the penetration completion state is greater than a preset target value, it is determined as the oil agent penetration completion state; In the output layer, the effective oil agent penetration amount is calculated according to the feature vector corresponding to the oil agent penetration completion state to obtain a quantitative penetration depth index characterizing the oil agent distribution ratio of each layer of the chenille yarn steaming fabric.

6. The intelligent monitoring method for the chemical fiber oil agent of chenille steamed yarn fabric according to claim 1, wherein Performing penetration and uniformity dual-task evaluation based on the quantitative penetration depth index to generate an oil agent deep penetration process parameter combination for the chenille yarn steaming process, including: Performing multi-point sampling on the chenille yarn steaming fabric based on the quantitative penetration depth index to obtain penetration depth distribution data in the fabric width direction; Inputting the quantitative penetration depth index into the shared encoding layer of the dual-task neural network to obtain an intermediate feature representation; Enhancing the attention mechanism of the intermediate feature representation through a feature interaction module to obtain an enhanced feature of information interaction between tasks; Inputting the enhanced feature into the penetration state decoding layer and the uniformity decoding layer respectively for dual decoding to obtain a penetration state evaluation value and a uniformity evaluation value; Calculating the effective penetration amount of the whole fabric according to the penetration state evaluation value, and calculating the coefficient of variation of the penetration amount in the fabric width direction according to the uniformity evaluation value to obtain an oil agent penetration quality comprehensive index; Generating an oil agent deep penetration process parameter combination for the chenille yarn steaming process according to the oil agent penetration quality comprehensive index.

7. The intelligent monitoring method for the chemical fiber oil agent of chenille yarn steaming fabric according to claim 6, characterized in that, Generating an oil agent deep penetration process parameter combination for the chenille yarn steaming process according to the oil agent penetration quality comprehensive index, including: Based on the comprehensive index of the oil agent penetration quality, establish the mapping relationship between the steam parameters and the oil agent penetration performance, and obtain the influence relationship model of the four process parameters of steam pressure, temperature, humidity and treatment time on the penetration effect; Conduct parameter sensitivity analysis on the influence relationship model, obtain the ranking of the influence degree of each parameter on the comprehensive index of the oil agent penetration quality, and determine the priority order and the best adjustment range of parameter adjustment through gradient calculation; Construct a fuzzy control rule base according to the expert knowledge and historical data of the chenille yarn steaming process, obtain the steam parameter adjustment strategy, and perform parameter adjustment calculation based on the fuzzy control rule base to obtain the parameter combination of the deep penetration process of the oil agent.

8. An intelligent monitoring system for chemical fiber oil agent of chenille steaming yarn fabric, characterized in that, For implementing the intelligent monitoring method for the chemical fiber oil agent of the chenille yarn steaming fabric as described in any one of claims 1-7, the intelligent monitoring system for the chemical fiber oil agent of the chenille yarn steaming fabric includes: An acquisition module, configured to perform multi-wavelength near-infrared spectroscopy and infrared thermal imaging acquisition on the chenille yarn steaming fabric during the yarn steaming process to obtain the monitoring data of the double-layer distribution of the oil agent; A transformation module, configured to perform spectral-temperature transformation on the monitoring data of the double-layer distribution of the oil agent to obtain the multi-layer penetration feature vector of the oil agent; A state analysis module, configured to input the multi-layer penetration feature vector of the oil agent into the oil agent penetration time series recognition network for state analysis to obtain the quantitative penetration depth index; A generation module, configured to perform dual-task evaluation of penetration and uniformity based on the quantitative penetration depth index, and generate the parameter combination of the deep penetration process of the oil agent for the chenille yarn steaming process.

Images