A digital analysis-based control method and system for a hemp fiber needle gill machine

Abstract

The present application relates to the technical field of hemp fiber production, and particularly relates to a hemp fiber needle carding machine control method and system based on digital analysis, which comprises the following steps: acquiring fiber tension, fiber distribution images and equipment operation parameters in real time through a tension sensor and an image acquisition device, and constructing a real-time acquisition dataset; extracting fiber length distribution features, fiber orientation features, fiber nodule features and fiber strength features to form a four-dimensional feature vector and input the four-dimensional feature vector into a support vector machine to predict quality indexes such as fiber separation degree, parallelism and damage rate; calculating a deviation vector according to a predicted value and a target value, and dynamically adjusting running parameters such as needle plate speed, carding angle and draft ratio through a fuzzy PID algorithm; obtaining actual quality parameters through online quality detection, comparing the actual quality parameters with predicted results to generate quality deviation data, and correcting equipment running parameters according to digital analysis. The present application realizes real-time optimization and stable control of hemp fiber carding quality.

Description

Technical Field

[0001] This invention relates to the field of hemp fiber production technology, specifically to a control method and system for a hemp fiber combing machine based on digital analysis. Background Technology

[0002] Hemp fiber is a typical long-pitch natural fiber, characterized by high toughness, uneven length, and high sizing content. Its combing process is a crucial step in hemp yarn production. The carding machine, as the core equipment in the hemp fiber processing flow, achieves fiber separation, parallelization, and impurity removal through the relative movement between the carding plates and the fibers. Its operating status and process parameters directly affect the fiber separation degree, parallelism, and damage rate. Therefore, accurately acquiring the status of the combing process and controlling the equipment operating parameters in real time are important technical means to improve the quality of hemp fibers and the spinning performance.

[0003] However, existing control methods for hemp fiber carding machines still rely primarily on manual parameter settings based on experience. This can easily lead to unstable carding quality due to variations in fiber batches, fluctuations in ambient temperature and humidity, or equipment wear. For example, improper settings for the carding speed or draft ratio can result in over-drafting or insufficient carding, increasing fiber damage. An unreasonable carding angle can cause a decrease in fiber parallelism, thus affecting the uniformity of subsequent spinning. Traditional control methods lack real-time detection and quantitative analysis of fiber conditions, making it difficult to achieve dynamic parameter optimization based on actual carding results.

[0004] Therefore, it is necessary to design a control method and system for hemp fiber combing machines based on digital analysis to solve the problems existing in the current technology. Summary of the Invention

[0005] In view of this, the present invention proposes a control method and system for a hemp fiber combing machine based on digital analysis, aiming to solve the problem that existing hemp fiber combing machines lack real-time detection and quantitative analysis of fiber state, making it difficult to achieve dynamic parameter optimization based on actual combing effect.

[0006] This invention proposes a control method for a hemp fiber combing machine based on digital analysis, comprising: Based on high-precision tension sensors and image acquisition devices, fiber tension data, fiber distribution image data and equipment operating parameters are collected in real time during the combing process of hemp fibers, and a real-time acquisition dataset is constructed. The fiber distribution image data is subjected to multi-scale wavelet transform processing to extract fiber length distribution features, fiber orientation features, and fiber nodule features; The fiber tension data is subjected to adaptive filtering and normalization to extract fiber strength features; A four-dimensional feature vector is formed based on the combination of fiber length distribution characteristics, fiber orientation characteristics, fiber nodule characteristics and fiber strength characteristics, and the fiber combing quality index is predicted by the support vector machine algorithm. The deviation vector is calculated based on the predicted fiber combing quality index and the preset target quality index, and the equipment operating parameters are dynamically adjusted according to the deviation vector using a fuzzy PID control algorithm. The quality parameter set of combed hemp fibers is monitored in real time through online quality detection. The quality parameter set includes fiber separation degree parameter, fiber parallelism parameter, and fiber damage rate parameter. The quality parameter set is compared with the predicted quality indicators to generate quality deviation data, and the equipment operating parameters are corrected based on digital analysis.

[0007] Furthermore, based on high-precision tension sensors and image acquisition devices, fiber tension data, fiber distribution image data, and equipment operating parameters are collected in real time during the hemp fiber combing process. When constructing the real-time acquisition dataset, it includes: A high-precision tension sensor array is arranged at the feed inlet, carding zone and discharge outlet of the hemp fiber carding area to acquire fiber tension data; During the combing process, fiber distribution image data is acquired at a fixed frequency using an image acquisition device. The needle plate movement speed, combing angle, draw ratio, and feed speed are obtained as equipment operating parameters; Timestamps are added to the collected fiber tension data, fiber distribution image data, and equipment operating parameters; and synchronization alignment is performed based on the timestamps; the synchronized and aligned fiber tension data, fiber distribution image data, and equipment operating parameters form the real-time acquisition dataset.

[0008] Furthermore, when performing multi-scale wavelet transform processing on the fiber distribution image data to extract fiber length distribution features, fiber orientation features, and fiber nodule features, the process includes: The fiber distribution image data were decomposed into three levels using the Daubechies wavelet basis function; the energy distribution of the wavelet coefficients at each scale was calculated to determine the fiber length distribution characteristics. Gabor filter banks were used to analyze the directional response of the image and extract fiber orientation features. Fiber nodule regions in images are detected through morphological processing and Hough transform; the area, density, and distribution uniformity of the nodule regions are calculated as fibrous nodule features.

[0009] Furthermore, when performing adaptive filtering and normalization processing on the fiber tension data to extract fiber strength features, the process includes: The fiber tension data is filtered using the minimum mean square error algorithm; the filtered fiber tension data is statistically analyzed to calculate the average tension value and tension fluctuation range; the initial fiber strength characteristics are determined based on the average tension value and tension fluctuation range; the initial fiber strength characteristics are normalized to finally form the fiber strength characteristics.

[0010] Furthermore, when predicting fiber combing quality indicators using the support vector machine algorithm, the following are included: The fiber length distribution characteristics, fiber orientation characteristics, fiber nodule characteristics, and fiber strength characteristics are combined to form a four-dimensional feature vector; the similarity between the four-dimensional feature vector and each support vector in the support vector machine is calculated based on the radial basis function kernel; each similarity is multiplied by the corresponding Lagrange multiplier and summed; a bias term is added to the summation result to obtain a preliminary predicted value; the preliminary predicted value is linearly transformed to map it to the physical range of fiber combing quality indicators; the predicted values ​​of fiber separation degree, fiber parallelism, and fiber damage rate are output.

[0011] Furthermore, when calculating the deviation vector based on the predicted fiber combing quality index and the preset target quality index, it includes: The preset target quality indicators include fiber separation target value, fiber parallelism target value, and fiber damage rate target value; The fiber separation deviation is obtained by subtracting the target fiber separation value from the predicted fiber separation value; the fiber parallelism deviation is obtained by subtracting the target fiber parallelism value from the predicted fiber parallelism value. The fiber damage rate deviation is obtained by subtracting the predicted fiber damage rate from the target fiber damage rate. The fiber separation deviation, fiber parallelism deviation, and fiber damage rate deviation are normalized; the normalized fiber separation deviation, fiber parallelism deviation, and fiber damage rate deviation are then organized into a deviation vector.

[0012] Furthermore, when dynamically adjusting the equipment operating parameters based on the deviation vector using a fuzzy PID control algorithm, the following steps are included: The deviation vector is fuzzified to generate a corresponding fuzzy set; fuzzy inference is performed based on the fuzzy set using a preset fuzzy rule base to generate a fuzzy inference result; the output is aggregated using the minimum-maximum synthesis method based on the fuzzy inference result to generate a fuzzy output; and defuzzification is performed using the centroid method to generate precise PID parameter adjustment values; the basic PID parameters are updated based on the PID parameter adjustment values ​​to generate adjusted PID parameters; the control quantity is calculated based on the adjusted PID parameters to generate a control quantity calculation result; the control quantity calculation result is converted into needle plate movement speed adjustment values, combing angle adjustment values, and draw ratio adjustment values, and the adjustment instructions for the equipment operating parameters are output.

[0013] Furthermore, when monitoring the quality parameter set of combed hemp fibers in real time through online quality inspection, the parameters include: Image data from the outlet of the combing machine is collected to generate a fiber distribution image. Based on the fiber distribution image, the fiber spacing uniformity is calculated using an image processing algorithm to generate a fiber separation parameter. The fiber orientation consistency is calculated using a direction analysis algorithm to generate a fiber parallelism parameter. Data on the breakage points of hemp fibers at the outlet of the combing machine are collected, and the number of breaks per unit length is calculated to generate a fiber damage rate parameter. Based on the fiber separation parameter, fiber parallelism parameter, and fiber damage rate parameter, a set of tissue formation quality parameters is established.

[0014] Furthermore, when correcting equipment operating parameters based on digital analysis, this includes: Based on the quality deviation data, the mapping relationship between the feature vector and the quality index is updated through an incremental learning algorithm, and the parameter compensation coefficient is calculated by combining the environmental temperature and humidity change data. Based on the parameter compensation coefficients, the correction values ​​for needle plate movement speed, combing angle, and draft ratio are generated; and based on the correction values ​​for needle plate movement speed, combing angle, and draft ratio, the corrected equipment operating parameters are output.

[0015] Compared with existing technologies, the advantages of this invention are as follows: By deploying tension sensor arrays at the feed inlet, carding zone, and discharge outlet, and acquiring fiber distribution images at a fixed frequency, combined with timestamp synchronization technology, multi-dimensional real-time acquisition of fiber tension changes, fiber spatial distribution, and equipment operating status can be achieved, providing a high-quality data foundation for subsequent precise control. Employing Daubechies wavelet, Gabor filter bank, and morphological-Hough combination methods, multiple key structural features such as fiber length, orientation, and knots can be captured simultaneously. Compared with traditional single-scale or single-feature methods, this improves image analysis accuracy and enhances the quantitative description of fiber state. Noise interference is eliminated through minimum mean square error adaptive filtering, and tension statistical features are normalized, ensuring stable consistency of fiber strength features under different environmental or equipment conditions, thereby improving the reliability of the overall prediction model. By constructing a combined feature vector from four types of features—length distribution, orientation, knots, and strength—real-time prediction of fiber separation, parallelism, and damage rate is achieved through RBF kernel SVM, enabling early identification of carding deviations, reducing quality fluctuations, and providing predictive control capabilities for the carding process. By fuzzifying the deviation vector and using the centroid method for defuzzification to generate dynamic PID parameters, precise control is maintained even under varying operating conditions, fiber state fluctuations, or load disturbances. Compared to fixed PID parameters, this reduces response overshoot and improves the stability of combing quality. Real-time detection of fiber separation, parallelism, and damage rate, compared with predicted results to generate quality deviation data, enables closed-loop correction of the control strategy, reducing output quality volatility and improving product consistency. Utilizing online quality data and changes in ambient temperature and humidity, the feature-quality mapping relationship is updated incrementally through learning, and parameter compensation coefficients are introduced. This allows for high control accuracy to be maintained even under long-term operation, changes in raw material batches, or seasonal temperature and humidity variations, achieving long-term adaptive optimization.

[0016] On the other hand, this application also provides a hemp fiber combing machine control system based on digital analysis, for applying the above-mentioned hemp fiber combing machine control method based on digital analysis, including: The acquisition unit is configured to acquire fiber tension data, fiber distribution image data and equipment operating parameters in real time during the hemp fiber combing process based on a high-precision tension sensor and image acquisition device, and construct a real-time acquisition dataset. The processing unit is configured to perform multi-scale wavelet transform processing on the fiber distribution image data to extract fiber length distribution features, fiber orientation features, and fiber nodule features; and to perform adaptive filtering processing on the fiber tension data and normalize the data to extract fiber strength features. The adjustment unit is configured to form a four-dimensional feature vector based on the combination of fiber length distribution characteristics, fiber orientation characteristics, fiber nodule characteristics and fiber strength characteristics, and predict fiber combing quality indicators through a support vector machine algorithm; calculate the deviation vector based on the predicted fiber combing quality indicators and the preset target quality indicators, and dynamically adjust the equipment operating parameters based on the deviation vector through a fuzzy PID control algorithm. The correction unit is configured to monitor the set of quality parameters of the combed hemp fibers in real time through online quality detection. The set of quality parameters includes fiber separation parameters, fiber parallelism parameters, and fiber damage rate parameters. The unit compares the set of quality parameters with the predicted quality indicators to generate quality deviation data and corrects the equipment operating parameters based on digital analysis.

[0017] It is understandable that the above-mentioned control method and system for hemp fiber combing machines based on digital analysis have the same beneficial effects, and will not be elaborated further here. Attached Figure Description

[0018] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A flowchart of a hemp fiber combing machine control method based on digital analysis provided in an embodiment of the present invention; Figure 2 This is a functional block diagram of a hemp fiber combing machine control system based on digital analysis, provided for an embodiment of the present invention. Detailed Implementation

[0019] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the disclosure to those skilled in the art. It should be noted that, unless otherwise specified, embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0020] For this, please refer to Figure 1 As shown, this application proposes a control method for a hemp fiber combing machine based on digital analysis, including: S100: Based on a high-precision tension sensor and image acquisition device, it collects fiber tension data, fiber distribution image data and equipment operating parameters in real time during the hemp fiber combing process, and constructs a real-time acquisition dataset; S200: Perform multi-scale wavelet transform processing on fiber distribution image data to extract fiber length distribution features, fiber orientation features, and fiber nodule features; S300: Adaptive filtering and normalization are performed on fiber tension data to extract fiber strength features; S400: A four-dimensional feature vector is formed by combining fiber length distribution characteristics, fiber orientation characteristics, fiber nodule characteristics and fiber strength characteristics, and the fiber combing quality index is predicted by the support vector machine algorithm. S500: Calculates the deviation vector based on the predicted fiber combing quality index and the preset target quality index, and dynamically adjusts the equipment operating parameters based on the deviation vector using a fuzzy PID control algorithm; S600: Real-time monitoring of the quality parameter set of combed hemp fibers through online quality detection. The quality parameter set includes fiber separation degree parameter, fiber parallelism parameter, and fiber damage rate parameter. The quality parameter set is compared with the predicted quality indicators to generate quality deviation data, and the equipment operating parameters are corrected based on digital analysis.

[0021] Specifically, the digital analysis-based control method for a hemp fiber combing machine proposed in this application achieves precise monitoring and intelligent control of the hemp fiber combing process through high-precision sensing technology and advanced digital signal processing algorithms. A high-precision tension sensor array is arranged at the feed inlet, combing zone, and discharge outlet of the hemp fiber combing area, and an industrial-grade high-definition image acquisition device is also provided to collect fiber tension data, fiber distribution image data, and equipment operating parameters in real time. Image processing employs multi-scale wavelet transform technology to accurately extract fiber length distribution features, fiber orientation features, and fiber nodule features; tension data is processed through an adaptive filtering algorithm to extract high-precision fiber strength features. These features are combined to form a multi-dimensional feature vector, which is then used to predict fiber combing quality indicators, including fiber separation, fiber parallelism, and fiber damage rate, through a trained support vector machine model. Based on the deviation between the predicted results and the target values, a fuzzy PID control algorithm is used to dynamically adjust key parameters such as the needle plate movement speed, combing angle, draft ratio, and feed speed, achieving real-time optimization of the combing process. In addition, it is equipped with an online quality inspection module to monitor the combed hemp fibers in real time, compare the actual quality parameters with the predicted values, and continuously correct the mapping relationship between feature vectors and quality indicators through incremental learning algorithms to form a closed-loop control and ensure continuous optimization of combing quality.

[0022] This method solves the problems of traditional hemp fiber carding processes, such as reliance on manual experience for quality control, delayed parameter adjustments, and large quality fluctuations. Through digital and intelligent control strategies, it achieves a shift from "experience-based control" to "precision control," improving hemp fiber carding quality and production efficiency. It can sense changes in fiber state in real time, predict carding quality trends in advance, and proactively adjust equipment parameters, avoiding the passive "problem-manual intervention-quality recovery" model of traditional methods. By moving the quality control point forward, it significantly reduces fiber damage and quality fluctuations.

[0023] The working process and principle of this application are as follows: a high-precision tension sensor array is arranged at the feed inlet, carding zone, and discharge outlet of the hemp fiber carding area to monitor the tension changes of the fibers in real time during the carding process. Simultaneously, an image acquisition device acquires fiber distribution image data at a fixed frequency to capture the fiber arrangement and structural characteristics. Equipment operating parameters such as needle plate movement speed, carding angle, draw ratio, and feed speed are also acquired synchronously to form a complete real-time acquisition dataset. For the acquired fiber distribution image data, multi-scale wavelet transform processing is performed: the image is decomposed into multiple layers using specific wavelet basis functions, and the energy distribution of wavelet coefficients at each scale is calculated. The fiber length distribution characteristics are determined through the energy distribution features. A direction-sensitive filter bank is applied to analyze the image response in different directions to extract fiber orientation features, which reflect the fiber arrangement consistency and orientation distribution. Fiber nodule regions in the image are detected through morphological processing and specific transformations. The area, density, and distribution uniformity of the nodule regions are calculated as fiber nodule features, which reflect the degree of fiber entanglement and carding difficulty. For fiber tension data, adaptive filtering is implemented: an adaptive filter is designed based on a specific algorithm, and the filtering parameters are dynamically adjusted according to the statistical characteristics of the tension signal to effectively remove measurement noise and external interference. Statistical analysis is performed on the filtered tension data to calculate the average tension value and tension fluctuation range, and the initial fiber strength characteristics are determined by combining the physical properties of hemp fibers. Finally, the initial characteristics are normalized to eliminate the absolute strength differences between different batches of hemp fibers, forming standardized fiber strength characteristics. Fiber length distribution characteristics, fiber orientation characteristics, fiber nodule characteristics, and fiber strength characteristics are combined to form a multi-dimensional feature vector, which is input into a pre-trained support vector machine model. This model calculates the similarity between the feature vector and the support vector based on a specific kernel function, obtains preliminary predicted values ​​through weighted summation and the addition of a bias term, and then maps these values ​​to the actual physical range through linear transformation, outputting predicted values ​​for fiber separation, fiber parallelism, and fiber damage rate. The predicted fiber combing quality indicators are compared with the preset target quality indicators to calculate the deviation vector: the predicted fiber separation value minus the target value yields the separation deviation; the predicted fiber parallelism value minus the target value yields the parallelism deviation; and the target fiber damage rate value minus the predicted value yields the damage rate deviation. These deviations are normalized and organized into a multi-dimensional deviation vector. Based on the deviation vector, a fuzzy PID control algorithm is used to dynamically adjust the equipment operating parameters: first, the deviation vector is fuzzified to generate a corresponding fuzzy set; then, inference operations are performed according to a preset fuzzy rule base to generate fuzzy inference results; a specific synthesis method is used for output aggregation to generate fuzzy output; a specific method is used for defuzzification to obtain precise PID parameter adjustment amounts; the basic PID parameters are updated based on the adjustment amounts, control quantities are calculated, and finally, adjustment instructions for equipment operating parameters such as needle plate movement speed adjustment, combing angle adjustment, and draft ratio adjustment are generated.It is also equipped with an online quality inspection module to monitor the quality of the combed hemp fibers in real time: It collects image data from the discharge port, calculates fiber spacing uniformity using image processing algorithms to generate fiber separation parameters, calculates fiber orientation consistency using direction analysis algorithms to generate fiber parallelism parameters, and generates fiber damage rate parameters by counting the number of breaks per unit length. These parameters are organized into a quality parameter set and compared with predicted quality indicators to generate quality deviation data. Finally, parameter corrections are implemented based on the quality deviation data: the mapping relationship between feature vectors and quality indicators is updated using incremental learning algorithms, and parameter compensation coefficients are calculated in conjunction with environmental temperature and humidity change data; correction values ​​for needle plate movement speed, combing angle, and draft ratio are generated based on the compensation coefficients; and the corrected equipment operating parameters are output, forming a closed-loop control to ensure continuous optimization of combing quality.

[0024] As a preferred embodiment, the solution of this application is implemented as follows: This control method is implemented on a flax fiber carding machine in a flax textile enterprise. The carding machine processes flax fibers with an average fiber length of 25-35mm and a fineness of 15-20tex. Six high-precision tension sensors (model: Zemic H3C3-50N, accuracy 0.01N) are installed at the feed inlet, carding zone, and discharge outlet of the carding zone, forming an 18-point tension monitoring network. The image acquisition device uses a Baslerac A2000-165um industrial camera, equipped with a 50mm fixed-focus lens and a ring LED light source, installed above the carding zone, to acquire fiber distribution images at a frame rate of 90fps. The basic values ​​for the needle plate movement speed are set to 1200rpm, the basic value for the carding angle is 15°, the basic value for the draft ratio is 1.8, and the basic value for the feed speed is 8m / min. The target quality indicators are set as follows: fiber separation ≥92%, fiber parallelism ≥88%, and fiber damage rate ≤5%. During operation, data is acquired every 50ms to form a real-time acquisition dataset. Three-layer Daubechies wavelet decomposition (db4 wavelet basis) was performed on the fiber distribution image to calculate the energy distribution at each scale and determine the fiber length distribution characteristics. Eight Gabor filters with different orientations (center frequency 0.15 cycles / pixel, directional spacing 22.5°) were applied to extract fiber orientation features. Fiber nodules were detected through morphological opening operations (structural element radius 3 pixels) and Hough transform, and the area ratio, density, and distribution entropy of the nodule region were calculated as fiber nodule features. Adaptive filtering based on the LMS algorithm (step size 0.005, filter order 32) was applied to the tension data to calculate the average tension value and standard deviation within a 100ms window, which were then normalized to form fiber strength features. The four-dimensional feature vector was input into a pre-trained support vector machine model (C=10, γ=0.5) to predict fiber combing quality indicators. When the predicted fiber separation is 89% (lower than the target 92%), fiber parallelism is 85% (lower than the target 88%), and fiber damage rate is 6.5% (higher than the target 5%), the deviation vector is calculated as follows: separation deviation -3%, parallelism deviation -3%, and damage rate deviation +1.5%. After normalization, the deviation vector is (-0.6, -0.6, 0.3). The fuzzy PID controller fuzzifies the deviation vector into "negative large", "negative medium", and "positive small", and infers the PID parameter adjustment based on the fuzzy rule base: proportional coefficient increases by 15%, integral time decreases by 10%, and derivative time increases by 5%. The updated PID parameters calculate the control quantity, generating adjustment instructions to increase the needle plate movement speed by 2%, decrease the combing angle by 1°, and increase the draft ratio by 0.05. The online quality inspection module collects images from the discharge port and calculates the actual fiber separation as 90.5%, parallelism as 86.2%, and damage rate as 5.8%, which deviates from the predicted values.The support vector machine model is updated using an incremental learning algorithm. Combined with the current ambient temperature and humidity (25℃, 60% humidity), parameter compensation coefficients are calculated: needle plate movement speed compensation +0.5%, combing angle compensation -0.2°, and draw ratio compensation +0.02. The final corrected equipment operating parameters are output. After 24 hours of continuous operation, fiber separation was stabilized at 93.5±0.8%, fiber parallelism at 89.2±0.7%, and fiber damage rate controlled at 4.3±0.5%, representing improvements of 3.5%, 2.2%, and a reduction of 1.7% respectively compared to traditional control methods.

[0025] Through the above-mentioned scheme, this application achieves comprehensive perception of fiber state and accurate prediction of quality indicators by using multi-source data fusion and advanced algorithms, thus moving the quality control point forward and avoiding the occurrence of quality problems. The fuzzy PID control algorithm enables dynamic optimization of equipment parameters, ensuring that the combing process is always in optimal condition. Through online quality detection and parameter correction mechanisms, a closed-loop control is formed, which can adapt to the differences in characteristics of different batches of hemp fibers and environmental changes, maintaining stable combing quality. This improves the separation and parallelism of hemp fibers, reduces fiber damage rate, and enhances the quality and added value of the final product.

[0026] This application further proposes a method based on a high-precision tension sensor and image acquisition device to collect fiber tension data, fiber distribution image data, and equipment operating parameters in real time during the hemp fiber combing process. When constructing the real-time acquisition dataset, it includes: A high-precision tension sensor array is arranged at the feed inlet, carding zone and discharge outlet of the hemp fiber carding area to acquire fiber tension data; During the combing process, fiber distribution image data is acquired at a fixed frequency using an image acquisition device. The needle plate movement speed, combing angle, draw ratio, and feed speed are obtained as equipment operating parameters; Timestamps are added to the collected fiber tension data, fiber distribution image data, and equipment operating parameters; and synchronization alignment is performed based on the timestamps; the synchronized and aligned fiber tension data, fiber distribution image data, and equipment operating parameters form a real-time acquisition dataset.

[0027] Specifically, the tension sensor array employs a distributed layout strategy, placing multiple sensor points at key locations within the combing area to cover important regions. Each tension sensor is designed for high precision, featuring high sensitivity and fast response, and is connected to the signal conditioning circuit via shielded twisted-pair cables. The signal conditioning circuit utilizes a multi-stage filtering architecture: a passive low-pass filter at the first stage, an active filter at the second stage, and a digital filter at the third stage, employing an adaptive filtering algorithm to dynamically adjust filtering parameters based on the real-time signal-to-noise ratio. After conditioning, the tension signal is digitized via a high-resolution analog-to-digital converter, using a high sampling rate to ensure the capture of rapidly changing tension signals. The image acquisition device uses an industrial-grade high-definition camera equipped with an appropriate lens and an adjustable light source. The light source includes LEDs of multiple wavelengths, allowing switching between different wavelengths to adapt to different fiber characteristics. The camera is connected to the image processing unit via a high-speed interface, employing a hardware trigger mode to ensure precise synchronization between image acquisition and equipment operation. The image acquisition frequency is dynamically adjusted according to the combing speed, automatically increasing the acquisition frequency when the combing speed changes significantly. Equipment operating parameters are acquired through equipment control: needle plate movement speed is measured by a high-precision encoder; combing angle is measured by an angle sensor; draw ratio is calculated by the front and rear roller speed ratio; and feeding speed is measured by a photoelectric encoder. These parameters are acquired in real time from equipment control via an industrial network, employing a high sampling rate to ensure the capture of detailed changes in equipment operation. The time synchronization mechanism uses a precision time protocol, with one master clock node and other data acquisition nodes acting as slave clocks. The master clock obtains standard time through a GPS receiver, and slave clocks synchronize with the master clock via the protocol, achieving high-precision time synchronization. Each data packet contains a precise timestamp, data quality markers, and verification information. Data synchronization alignment employs dynamic time warping technology to address the issue of inconsistent sampling frequencies from different data sources, ensuring precise data correspondence in the time dimension. Data integrity verification is implemented using cyclic redundancy check and data packet sequence number checking; when data loss or errors are detected, a retransmission mechanism is automatically triggered. The data acquisition process also implements an adaptive sampling strategy, dynamically adjusting the sampling frequency according to the state, increasing the sampling rate during rapid state changes and decreasing the sampling rate during steady-state operation to optimize resource utilization.

[0028] Through the above technical solutions, this application achieves comprehensive and accurate monitoring of the hemp fiber combing process; and through high-precision time synchronization technology, it ensures the time consistency of multi-source heterogeneous data.

[0029] This application further proposes a method for extracting fiber length distribution features, fiber orientation features, and fiber nodule features by performing multi-scale wavelet transform processing on fiber distribution image data, including: A multi-level wavelet decomposition of fiber distribution image data is performed using a specific wavelet basis function; the energy distribution of wavelet coefficients at each scale is calculated to determine the fiber length distribution characteristics. A direction-sensitive filter bank is used to analyze the directional response of the image and extract fiber orientation features; Fiber nodule regions in images are detected through morphological processing and specific transformations; the area, density, and distribution uniformity of the nodule regions are calculated as fibrous nodule features.

[0030] Specifically, multi-scale wavelet transform processing employs specific wavelet basis functions to perform multi-level wavelet decomposition on the fiber distribution image. The decomposition process first performs a one-dimensional wavelet transform on the original image to obtain low-frequency approximation coefficients and high-frequency detail coefficients; then, a second one-dimensional wavelet transform is performed on the low-frequency approximation coefficients to obtain the second-level decomposition result; finally, a third one-dimensional wavelet transform is performed on the low-frequency approximation coefficients of the second level to obtain the third-level decomposition result. To maintain image size consistency, upsampling and filtering are performed after each decomposition level to ensure that the coefficient size of each level is the same as the original image. The extraction of fiber length distribution features uses energy distribution analysis: the energy of high-frequency coefficients in the horizontal, vertical, and diagonal directions at each scale is calculated, and a mapping relationship between energy distribution and fiber length is established based on the physical properties of hemp fibers. Through normalization, the fiber proportion distribution in each length interval is obtained, forming a fiber length distribution feature vector. The extraction of fiber orientation features uses direction-sensitive filter bank analysis: multiple filters in different directions are designed with uniform directional spacing, and the center frequency and bandwidth are optimized. The original image is sequentially applied with filters in each direction, and the energy response of the filtered image is calculated. By analyzing the distribution of energy response, the main orientation angle and orientation consistency index are determined. Orientation histogram analysis is also applied to calculate orientation entropy, reflecting the degree of disorder in fiber orientation. Fiber nodule feature extraction employs a combination of morphological processing and specific transformations: first, the image is normalized and contrast enhanced; then, morphological opening operations are applied to remove small noise points; and finally, closing operations are applied to fill internal holes in the regions. Fiber regions are extracted through threshold segmentation, and local density is calculated. Regions with local density exceeding the threshold are marked as potential nodule regions. Specific transformations are applied to detect circular nodules, and the nodule center and radius are determined through parameter spatial accumulation. Finally, the proportion of the total area of ​​the nodule region to the total area of ​​the fiber region is calculated as the nodule area feature, the number of nodules per unit area is calculated as the nodule density feature, and the spatial distribution statistics of nodule locations are calculated as the distribution uniformity feature. Feature quality assessment is implemented, calculating confidence weights for each feature parameter and dynamically adjusting them based on image quality, fiber density, and environmental conditions. The feature extraction process also considers the effects of fiber overlap and crossing, applying deep learning-assisted recognition to improve the accuracy and robustness of feature extraction.

[0031] Through the above technical solution, this application achieves in-depth analysis of fiber distribution images and extracts key features reflecting fiber state.

[0032] This application further proposes methods for adaptive filtering and normalization of fiber tension data to extract fiber strength features, including: The fiber tension data is filtered using a specific algorithm; the filtered fiber tension data is statistically analyzed to calculate the average tension value and tension fluctuation range; the initial fiber strength characteristics are determined based on the average tension value and tension fluctuation range; the initial fiber strength characteristics are normalized to finally form the fiber strength characteristics.

[0033] Specifically, adaptive filtering employs an adaptive filter based on a specific algorithm. A transverse filter structure is constructed, and the filter parameters are dynamically adjusted according to the statistical characteristics of the input signal. The reference signal for the filter is an input signal delayed by one sample, enabling prediction of future tension. To improve filtering performance, a variable step-size algorithm is implemented: when the error signal is large, the step-size parameter is increased to accelerate convergence; when the error signal is small, the step-size parameter is decreased to reduce steady-state error. A dynamic adjustment strategy for the filter order is also applied, automatically adjusting the filter order according to the complexity of the input signal. When the signal is simple, the order is reduced to decrease computation; when the signal is complex, the order is increased to ensure filtering performance. Statistical analysis is performed on the filtered tension data: the average tension value and tension fluctuation range within a specific time window are calculated to reflect the stability and variation characteristics of the tension. The tension change rate is also calculated to reflect the dynamic variation characteristics of the tension. The initial fiber strength characteristics are determined considering the physical properties of hemp fibers: a mechanical model is applied to convert the average tension value into an estimated fiber strength value. The conversion coefficients are calibrated experimentally, taking into account the cross-sectional shape and arrangement of the fibers. The impact of tension fluctuation range on fiber strength is also considered, defining a strength stability index to reflect the uniformity of stress on the fiber during the combing process. Normalization employs a dynamic range normalization method: the dynamic range of fiber strength characteristics is determined based on historical data, and a normalization formula is applied. To accommodate the differences in characteristics between different batches of hemp fibers, adaptive normalization is implemented, dynamically updating the normalization range. Other normalization methods are also used as supplementary methods to handle outlier cases. Feature quality assessment is implemented, calculating confidence weights for fiber strength characteristics: the reliability score of the features is dynamically adjusted based on sensor status, environmental conditions, and signal quality indicators. When the confidence score falls below a threshold, it automatically switches to a backup feature extraction method to ensure the continuity and reliability of the feature data. The feature extraction process also considers the detection of fiber breakage events, identifying fiber breakage points through tension mutation analysis, providing a basis for predicting fiber damage rates.

[0034] Through the above technical solution, this application achieves high-precision processing of fiber tension data and extracts key features reflecting the fiber strength state.

[0035] This application further proposes methods for predicting fiber combing quality indicators using support vector machine algorithms, including: The fiber length distribution characteristics, fiber orientation characteristics, fiber nodule characteristics, and fiber strength characteristics are combined to form a multidimensional feature vector; the similarity between the multidimensional feature vector and each support vector in the support vector machine is calculated according to a specific kernel function; each similarity is multiplied by the corresponding Lagrange multiplier and summed; a bias term is added to the summation result to obtain the preliminary prediction value; the preliminary prediction value is linearly transformed to map it to the physical range of fiber combing quality indicators; the predicted values ​​of fiber separation degree, fiber parallelism, and fiber damage rate are output.

[0036] Specifically, the feature vector construction process employs feature fusion technology: combining features from multiple dimensions to form basic feature vectors. Principal component analysis is applied for feature dimensionality reduction, retaining the most representative feature combinations. Feature weighting is also implemented, dynamically adjusting weights based on the correlation between features and quality indicators to improve the predictive power of the feature vectors. The support vector machine model uses a specific kernel function, determining the optimal parameters through grid search and cross-validation. A multi-output support vector machine is implemented to predict multiple quality indicators simultaneously. The optimization problem is solved using a specific algorithm to obtain the optimal decision function. The prediction process involves multi-step computation: first, the kernel similarity between the input feature vector and each support vector is calculated; then, the similarity is multiplied by the corresponding Lagrange multipliers and summed to obtain the decision function value; next, the decision function value is linearly transformed to map to the physical range of the quality indicators; finally, physical constraints are applied to the predicted values ​​to ensure they remain within a reasonable range. Adaptive model updates are implemented: the support vector machine model is periodically incrementally learned using newly collected data, updating the support vectors and Lagrange multipliers. Model validation is implemented, using cross-validation to assess the model's predictive accuracy. Model retraining is triggered when the prediction error exceeds a threshold. Model parameters are dynamically adjusted, automatically selecting the optimal parameter combination based on environmental conditions and fiber characteristics. Uncertainty assessment of prediction results is performed: confidence intervals for predicted values ​​are calculated based on the support vector machine's distance function and historical prediction error distribution. Confidence intervals are used to evaluate the reliability of prediction results; if a confidence interval is too wide, additional data is automatically requested or a backup prediction model is switched to. Anomaly detection is also implemented to identify potential prediction errors and avoid parameter adjustments based on erroneous predictions. A multi-model fusion strategy is implemented: in addition to the main support vector machine model, a backup model is maintained as a supplement. When the confidence level of the main model's prediction is low, the system automatically switches to the backup model. The prediction results of the two models are fused through a weighted average, with weights dynamically adjusted based on the models' historical performance. This fused prediction improves robustness and adaptability, ensuring reliable prediction results under various conditions.

[0037] Through the above technical solutions, this application achieves high-precision prediction of fiber combing quality indicators; through model adaptation and uncertainty assessment, the reliability and applicability of the prediction results are ensured, and the accuracy and adaptability of control are improved.

[0038] This application further proposes a method for calculating the deviation vector based on the predicted fiber combing quality index and the preset target quality index, including: The preset target quality indicators include the target values ​​for fiber separation, fiber parallelism, and fiber damage rate. The fiber separation deviation is obtained by subtracting the target value from the predicted fiber separation value; the fiber parallelism deviation is obtained by subtracting the target value from the predicted fiber parallelism value; and the fiber damage rate deviation is obtained by subtracting the predicted value from the target value. These deviations are then normalized. Finally, the normalized deviations are organized into a deviation vector.

[0039] Specifically, preset target quality indicators are dynamically set based on product requirements and process specifications: fiber separation target values ​​are typically set in a high range, with the specific value depending on the final product's application; fiber parallelism target values ​​are typically set in a medium-to-high range, with higher parallelism required for high-count yarns; and fiber damage rate target values ​​are typically set in a low range, with lower damage rates required for finer products. Adaptive adjustment of target values ​​is implemented, dynamically optimizing them based on historical data and quality feedback to balance quality and efficiency. The deviation calculation process considers the physical characteristics and importance of quality indicators: for fiber separation and fiber parallelism, direct difference calculation is used, with negative values ​​indicating failure to meet the target; for fiber damage rate, inverse difference calculation is used, as lower damage rates are better, with positive values ​​indicating performance better than the target. Weighted deviation calculation is also implemented, assigning different weights based on the degree of influence of quality indicators on the final product quality; the weighted deviation reflects the overall quality status. Normalization uses a dynamic range normalization method: the typical variation range of each quality indicator is determined based on historical data, and a normalization formula is applied. To adapt to different process conditions, adaptive normalization is implemented, dynamically updating the normalization range. Other normalization methods are also applied as supplementary measures to handle abnormal deviations. The deviation vector organization implements a multi-level structure: the base layer is the original normalized deviation value, the intermediate layer is the weighted deviation value, and the advanced layer is the comprehensive deviation index. Deviation decomposition is implemented, breaking down the total deviation into steady-state deviation and dynamic deviation, used for proportional and derivative control respectively. Steady-state deviation is extracted using low-pass filtering, and dynamic deviation is extracted using high-pass filtering. Deviation trend analysis is also implemented to predict the future direction and magnitude of deviation changes, used for feedforward control. Deviation vector quality assessment implements multi-level checks: primary checks are based on statistical methods to identify deviations that significantly deviate from the normal range; intermediate checks are based on physical models to check whether the deviation conforms to the physical laws of the fiber combing process; advanced checks are based on historical data patterns to identify potential abnormal deviation patterns. When an abnormal deviation is detected, a diagnostic program is automatically initiated to determine possible causes and take corresponding measures. Deviation priority management is implemented: the processing priority is determined based on the degree of impact of the deviation on the final product quality. Fiber separation deviation usually has the highest priority, followed by fiber parallelism deviation, and finally fiber damage rate deviation. Priority management considers product specification requirements and customer feedback, dynamically adjusting the deviation handling strategy. It also implements deviation tolerance management, allowing a certain degree of tolerance for small fluctuations to avoid instability caused by excessive adjustments.

[0040] Through the above technical solutions, this application achieves accurate quantification and comprehensive evaluation of quality deviations; through deviation decomposition and priority management, it ensures the pertinence and effectiveness of parameter adjustments.

[0041] This application further proposes a method for dynamically adjusting equipment operating parameters based on the deviation vector using a fuzzy PID control algorithm, including: The deviation vector is fuzzified to generate a corresponding fuzzy set. Based on the fuzzy set, inference operations are performed using a pre-defined fuzzy rule base to generate fuzzy inference results. Based on the fuzzy inference results, a specific synthesis method is used to aggregate the output, generating fuzzy output. Then, a specific method is used to defuzzify the output, generating precise PID parameter adjustment values. Based on the PID parameter adjustment values, the basic PID parameters are updated to generate adjusted PID parameters. Based on the adjusted PID parameters, the control quantity is calculated, generating the control quantity calculation results. Based on the control quantity calculation results, these are converted into needle plate movement speed adjustment values, combing angle adjustment values, and draft ratio adjustment values, and the adjustment instructions for the equipment operating parameters are output.

[0042] Specifically, the fuzzification process employs an adaptive membership function generation technique: each component of the deviation vector is divided into multiple fuzzy sets, and the shape parameters of the membership function are dynamically adjusted based on historical data. Online clustering analysis is implemented, clustering historical deviation data into multiple clusters, with the center and dispersion of each cluster used to determine the parameters of the membership function. To handle the non-stationary nature of the deviation, a sliding window technique is applied, using only data from the most recent period for clustering analysis to ensure that the membership function can adapt to environmental changes. The boundaries of the fuzzy sets adopt a soft boundary design, with the membership functions of adjacent fuzzy sets partially overlapping to ensure a smooth transition of deviations in the boundary region. The fuzzy rule base adopts a hierarchical structure: the base layer contains core rules describing the basic laws of the fiber combing process; the optimization layer contains extended rules based on historical operational experience, extracted from historical data through association rule mining; and the adaptive layer contains dynamically generated rules, adjusted according to real-time performance. Each rule is equipped with a dynamic weight coefficient, reflecting the reliability and applicability of the rule, which is updated online through a reinforcement learning algorithm. Rule inference employs a weighted inference method, considering not only the matching degree between the rule's antecedents and inputs but also the confidence level of the rule itself, generating more reliable inference results. A rule conflict detection and resolution mechanism is implemented. When multiple rules produce conflicting conclusions, evidence theory is applied for fusion. Output aggregation employs a specific synthesis method: for each output variable, the output fuzzy set of all rules is calculated, and aggregation is performed by taking the maximum membership degree. To handle rule weights, weighted synthesis is implemented, with rule weights reflecting the reliability and applicability of the rules. A fuzzy quantifier-guided aggregation method is also applied, adjusting the aggregation strategy according to the urgency of the control objective, with a more aggressive aggregation strategy used in emergency situations. Defuzzification uses an improved centroid method: first, the centroid position of each output fuzzy set is calculated; then, confidence weights are applied to weight the centroid positions; finally, a nonlinear transformation maps the continuous output to actual physical quantities. To avoid the influence of extreme values ​​on the results, output smoothing is implemented, and the current output is weighted and averaged with historical outputs. Constraint processing is also applied, considering physical limitations during defuzzification to ensure that the output is within a feasible range. An adaptive mechanism is implemented for PID parameter adjustment: PID parameter adjustment amounts are generated based on fuzzy inference results. The basic PID parameters are set according to process specifications, and the adjusted PID parameters are the original parameters plus the adjustment amount. Parameter constraints are implemented to ensure that the adjusted PID parameters are within a reasonable range, avoiding instability. The PID parameters also consider equipment and fiber characteristics, applying gain scheduling technology to dynamically adjust parameters based on the operating point. The control quantity is calculated using an incremental PID algorithm: the difference between the current and previous control quantities is calculated, considering the current deviation, historical deviation, and rate of change of deviation. Anti-integral saturation measures are implemented, pausing the accumulation of the integral term when the control quantity reaches its limit. To handle multivariable coupling, a decoupling control strategy is applied, considering the mutual influence between parameters and avoiding mutual interference in parameter adjustments.Parameter adjustment is subject to safety monitoring: Real-time anomaly detection algorithms are applied to monitor for abnormal situations during parameter adjustment. When an anomaly is detected, safety protection procedures are automatically activated, including parameter rollback, degradation, and emergency shutdown. Safety monitoring also considers state continuity, performing state-space analysis to predict potential safety risks. Multi-layered security protection is implemented, including software protection, hardware protection, and manual intervention, to ensure safe operation.

[0043] Through the above technical solutions, this application achieves precise adjustment of equipment operating parameters and can dynamically optimize the combing process based on quality deviations; through fuzzy logic and adaptive mechanisms, it handles uncertainties, improves the robustness and adaptability of control, and ensures the stability and consistency of combing quality.

[0044] This application further proposes a method for real-time monitoring of a set of quality parameters for combed hemp fibers through online quality inspection, including: Image data from the outlet of the combing machine is collected to generate a fiber distribution image. Based on the fiber distribution image, the fiber spacing uniformity is calculated using an image processing algorithm to generate a fiber separation parameter. A direction analysis algorithm is used to calculate the fiber orientation consistency to generate a fiber parallelism parameter. Data on the breakage points of hemp fibers at the outlet of the combing machine are collected, and the number of breaks per unit length is calculated to generate a fiber damage rate parameter. Based on the fiber separation parameter, fiber parallelism parameter, and fiber damage rate parameter, a set of quality parameters is formed.

[0045] Specifically, the online quality inspection module employs a high-resolution industrial camera installed at an appropriate distance downstream of the combing machine's outlet, equipped with a ring LED light source and a polarizing filter to eliminate surface reflection interference. The camera continuously acquires images at an appropriate frame rate, averaging multiple frames at a time to reduce random noise. Automatic exposure control is implemented, dynamically adjusting exposure parameters based on image brightness to ensure stable image quality. To adapt to different fiber characteristics, multiple light source modes are provided, automatically selecting the optimal lighting conditions based on fiber type. The fiber separation parameter is calculated using a multi-scale analysis method: first, grayscale normalization and contrast enhancement are performed on the acquired images; then, multi-scale edge detection is applied to extract fiber boundaries; next, local fiber density is calculated; finally, the fiber spacing uniformity index is calculated. Spectral analysis technology is also applied to calculate spectral entropy as a supplementary indicator. The fiber separation parameter is a weighted combination of multiple indicators, with weights dynamically adjusted according to product requirements. The fiber parallelism parameter is calculated using orientation field analysis: First, an orientation-sensitive filter bank is applied to analyze the orientation response of the image; then, the principal orientation and orientation consistency of each pixel are calculated; next, a region growing algorithm is applied to cluster pixels with similar orientations into fiber bundles; finally, the standard deviation of the fiber bundle orientation is calculated as a parallelism index. A specific transformation is also applied to detect the overall fiber orientation, and the peak width of the orientation histogram is calculated as a supplementary index. The parallelism parameter is a weighted average of multiple indices, with the weights dynamically adjusted according to fiber type and product requirements. The fiber damage rate parameter is calculated using break point detection technology: two methods are implemented: image-based break detection and indirect measurement based on mechanical properties. The results of the two methods are fused to obtain the final fiber damage rate parameter. Time series analysis is also performed to calculate the trend of damage rate changes and predict future damage risk. The quality parameter set is organized using a hierarchical data structure: the base layer stores the original measurement values ​​and timestamps; the intermediate layer stores the calculated quality parameters; and the high-level layer stores the quality assessment results, including deviations from target values ​​and quality grades. Data quality assessment is implemented, calculating confidence weights for each quality parameter and dynamically adjusting based on image quality, sensor status, and environmental conditions. When the confidence level falls below a threshold, a remeasurement is automatically requested or a switch to an alternative measurement method is initiated. Quality trend analysis is also implemented: time series models are applied to analyze the changing trends of quality parameters and predict future quality status. Trend analysis considers both periodic and external factors to improve prediction accuracy. Anomaly detection is further implemented, identifying quality anomalies through statistical process control methods and triggering an early warning mechanism.

[0046] Through the above technical solutions, this application achieves comprehensive and accurate monitoring of the quality of combed hemp fibers; through multi-source data fusion and trend analysis, it improves the accuracy and foresight of quality assessment.

[0047] This application further proposes that when correcting equipment operating parameters based on digital analysis, the following should be included: Based on the quality deviation data, the mapping relationship between the feature vector and the quality index is updated through an incremental learning algorithm, and the parameter compensation coefficient is calculated by combining the environmental temperature and humidity change data. Based on the parameter compensation coefficient, the correction values ​​for the needle plate movement speed, combing angle, and draft ratio are generated. Based on the correction values ​​for the needle plate movement speed, combing angle, and draft ratio, the corrected equipment operating parameters are output.

[0048] Specifically, the quality deviation data analysis employs a multi-scale approach: decomposing the quality deviation data into three components: long-term trend, periodic fluctuations, and random noise. The long-term trend reflects slow performance changes and is extracted using moving average filtering; periodic fluctuations reflect regular changes and are extracted using spectral analysis; random noise reflects measurement errors and random interference and is eliminated using wavelet thresholding. Deviation source analysis is implemented, using causal inference techniques to identify the main factors causing quality deviations, distinguishing the influence of equipment parameters, fiber characteristics, and environmental conditions. The mapping relationship between feature vectors and quality indicators is updated using an incremental learning algorithm: applying an online learning algorithm to gradually update model parameters as new data arrives. Specific techniques are implemented to protect important parameters from excessive modification by new data, avoiding catastrophic amnesia. The model update process considers data quality, assigning higher weights to high-quality data and lower weights to low-quality data. Concept drift detection is also implemented; when a significant change in data distribution is detected, model retraining is triggered. Parameter compensation coefficient calculation employs multi-factor analysis: constructing a compensation coefficient calculation model with inputs including quality deviation data, environmental temperature and humidity variation data, fiber characteristic data, and equipment status data. The model is implemented using a neural network, comprising multiple processing layers and employing appropriate activation functions. Feature importance analysis is performed to determine the degree of influence of each factor on the compensation coefficient, and the model structure is dynamically adjusted. The compensation coefficient calculation also considers historical compensation effects, and the compensation strategy is optimized through reinforcement learning to maximize the compensation effect. Parameter correction implements a hierarchical strategy: parameter correction is divided into short-term correction and long-term correction. Short-term correction targets immediate quality deviations, quickly adjusting equipment parameters; long-term correction targets slow performance changes, gradually optimizing basic parameters. Short-term correction uses proportional-integral control for rapid response to quality deviations; long-term correction uses adaptive learning for slow adjustment of the baseline. Correction amplitude limits are also implemented to avoid instability caused by parameter abrupt changes, and the correction amplitude is dynamically adjusted according to the current state. The calculation of the needle plate movement speed correction value considers fiber characteristics and the carding stage: for long fibers and high separation requirements, the speed is appropriately reduced to reduce fiber damage; for short fibers and high parallelism requirements, the speed is appropriately increased to enhance the carding effect. The impact of speed changes on subsequent processes is also considered, and coordinated control is implemented to avoid production interruptions caused by sudden speed changes. The carding angle correction value is calculated considering fiber orientation and nodule density: when fiber orientation consistency is low or nodule density is high, the carding angle is increased to enhance carding strength; when fiber damage rate is high, the carding angle is decreased to reduce carding strength. Fuzzy rules are also applied to dynamically adjust the correction strategy based on fiber condition. The draw ratio correction value is calculated considering fiber separation and parallelism: when fiber separation is low, the draw ratio is decreased to reduce fiber tension; when fiber parallelism is low, the draw ratio is increased to enhance fiber alignment. The impact of draw ratio on fiber damage is also considered, and safety constraints are implemented to ensure the draw ratio remains within a reasonable range. Parameter corrections undergo closed-loop verification: before implementing parameter corrections, the correction effect is predicted using a digital twin model, and potential risks are assessed.The predicted results are compared with the target quality indicators to calculate the expected quality improvement. When the expected improvement is insufficient or there is a risk, the correction strategy is automatically adjusted. After parameter correction, quality changes are continuously monitored to evaluate the actual correction effect, forming a closed-loop feedback.

[0049] The compensation coefficient calculation model, digital twin model, and neural network together constitute a closed-loop intelligent quality control system. The overall architecture includes four modules: a multi-source data acquisition layer, a feature processing and encoding layer, a core calculation and prediction layer, and an output and closed-loop optimization layer. In the multi-source data acquisition layer, quality deviation data (long-term trends, periodic fluctuations, and random noise components), environmental temperature and humidity changes, fiber characteristic parameters, and equipment status data are acquired in real time and cleaned, normalized, weighted, and synchronized to ensure the accuracy and usability of the input data. In the feature processing and encoding layer, feature selection, feature importance analysis, and dynamic feature weighting are performed on the acquired data to extract key features that significantly affect the compensation coefficient and quality improvement. Simultaneously, the input features are encoded and mapped to provide standardized vectors for neural network calculations. The core computation and prediction layer comprises a multi-layer neural network and a digital twin model: the neural network employs a multi-layer fully connected or convolutional structure combined with appropriate activation functions to achieve multi-factor mapping and prediction of compensation coefficients. It updates model parameters through incremental online learning and concept drift detection, protecting critical parameters from catastrophic amnesia. The digital twin model constructs a high-fidelity virtual mapping of production equipment and processes, simulating needle plate movement, combing angle, draft ratio, and fiber behavior. It performs simulation prediction and risk assessment on the compensation coefficients output by the neural network. The output and closed-loop optimization layer compares the compensation coefficients calculated by the neural network with the digital twin simulation results to evaluate expected quality improvements. If potential risks or insufficient improvements exist, the compensation strategy is automatically adjusted. Simultaneously, it dynamically corrects equipment parameters by combining short-term PI control and long-term adaptive learning, continuously monitoring actual quality indicators. Through closed-loop feedback, it continuously optimizes neural network weights, compensation coefficients, and the digital twin model, achieving safe, stable, and high-precision quality control of the production process.

[0050] Through the above technical solutions, this application achieves precise correction of equipment operating parameters and can dynamically optimize the process based on quality deviations and environmental changes; through incremental learning and closed-loop verification, it ensures the accuracy and safety of parameter correction and improves the adaptive capability and long-term stability of control.

[0051] In another preferred embodiment based on the above embodiments, see [reference] Figure 2 As shown, this invention also proposes a control system for a hemp fiber combing machine based on digital analysis, which is used to apply a control method for a hemp fiber combing machine based on digital analysis, including: The acquisition unit is configured to acquire fiber tension data, fiber distribution image data and equipment operating parameters in real time during the hemp fiber combing process based on a high-precision tension sensor and image acquisition device, and construct a real-time acquisition dataset. The processing unit is configured to perform multi-scale wavelet transform processing on fiber distribution image data to extract fiber length distribution features, fiber orientation features, and fiber nodule features; and to perform adaptive filtering processing on fiber tension data and normalize the data to extract fiber strength features. The adjustment unit is configured to form a four-dimensional feature vector based on a combination of fiber length distribution characteristics, fiber orientation characteristics, fiber nodule characteristics, and fiber strength characteristics, and to predict fiber combing quality indicators using a support vector machine algorithm; it calculates a deviation vector based on the predicted fiber combing quality indicators and a preset target quality indicator, and dynamically adjusts the equipment operating parameters based on the deviation vector using a fuzzy PID control algorithm. The correction unit is configured to monitor the set of quality parameters of the combed hemp fibers in real time through online quality detection. The set of quality parameters includes fiber separation parameters, fiber parallelism parameters, and fiber damage rate parameters. The unit compares the set of quality parameters with the predicted quality indicators, generates quality deviation data, and corrects the equipment operating parameters based on digital analysis.

[0052] In summary, by deploying tension sensor arrays at the feed inlet, carding zone, and discharge outlet, and acquiring fiber distribution images at a fixed frequency, combined with timestamp synchronization technology, multi-dimensional real-time acquisition of fiber tension changes, fiber spatial distribution, and equipment operating status can be achieved, providing a high-quality data foundation for subsequent precise control. Employing Daubechies wavelet, Gabor filter bank, and morphological-Hough combination methods, multiple key structural features such as fiber length, orientation, and knots can be captured simultaneously. Compared to traditional single-scale or single-feature methods, this improves image analysis accuracy and enhances the quantitative description of fiber state. Adaptive filtering using minimum mean square error eliminates noise interference, and normalization of tension statistical features ensures stable consistency of fiber strength characteristics under different environmental or equipment conditions, thereby improving the reliability of the overall prediction model. Constructing a combined feature vector from four types of features—length distribution, orientation, knots, and strength—and using RBF kernel SVM, real-time prediction of fiber separation, parallelism, and damage rate can be achieved. This allows for early identification of carding deviations, reduction of quality fluctuations, and predictive control capabilities in the carding process. By fuzzifying the deviation vector and using the centroid method for defuzzification to generate dynamic PID parameters, precise control is maintained even under varying operating conditions, fiber state fluctuations, or load disturbances. Compared to fixed PID parameters, this reduces response overshoot and improves the stability of combing quality. Real-time detection of fiber separation, parallelism, and damage rate, compared with predicted results to generate quality deviation data, enables closed-loop correction of the control strategy, reducing output quality volatility and improving product consistency. Utilizing online quality data and changes in ambient temperature and humidity, the feature-quality mapping relationship is updated incrementally through learning, and parameter compensation coefficients are introduced. This allows for high control accuracy to be maintained even under long-term operation, changes in raw material batches, or seasonal temperature and humidity variations, achieving long-term adaptive optimization.

[0053] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A control method for a hemp fiber combing machine based on digital analysis, characterized in that, include: Based on high-precision tension sensors and image acquisition devices, fiber tension data, fiber distribution image data and equipment operating parameters are collected in real time during the combing process of hemp fibers, and a real-time acquisition dataset is constructed. The fiber distribution image data is subjected to multi-scale wavelet transform processing to extract fiber length distribution features, fiber orientation features, and fiber nodule features; The fiber tension data is subjected to adaptive filtering and normalization to extract fiber strength features; A four-dimensional feature vector is formed based on the combination of fiber length distribution characteristics, fiber orientation characteristics, fiber nodule characteristics and fiber strength characteristics, and the fiber combing quality index is predicted by the support vector machine algorithm. The deviation vector is calculated based on the predicted fiber combing quality index and the preset target quality index, and the equipment operating parameters are dynamically adjusted according to the deviation vector using a fuzzy PID control algorithm. The quality parameter set of combed hemp fibers is monitored in real time through online quality detection. The quality parameter set includes fiber separation degree parameter, fiber parallelism parameter, and fiber damage rate parameter. The quality parameter set is compared with the predicted quality indicators to generate quality deviation data, and the equipment operating parameters are corrected based on digital analysis.

2. The control method for a hemp fiber combing machine based on digital analysis according to claim 1, characterized in that, Based on a high-precision tension sensor and image acquisition device, fiber tension data, fiber distribution image data, and equipment operating parameters are collected in real time during the hemp fiber combing process. The real-time acquisition dataset includes: A high-precision tension sensor array is arranged at the feed inlet, carding zone and discharge outlet of the hemp fiber carding area to acquire fiber tension data; During the combing process, fiber distribution image data is acquired at a fixed frequency using an image acquisition device. The needle plate movement speed, combing angle, draw ratio, and feed speed are obtained as equipment operating parameters; Timestamps are added to the collected fiber tension data, fiber distribution image data, and equipment operating parameters; and synchronization alignment is performed based on the timestamps; the synchronized and aligned fiber tension data, fiber distribution image data, and equipment operating parameters form the real-time acquisition dataset.

3. The control method for a hemp fiber combing machine based on digital analysis according to claim 2, characterized in that, When performing multi-scale wavelet transform processing on the fiber distribution image data to extract fiber length distribution features, fiber orientation features, and fiber nodule features, the process includes: The fiber distribution image data were decomposed into three levels using the Daubechies wavelet basis function; the energy distribution of the wavelet coefficients at each scale was calculated to determine the fiber length distribution characteristics. Gabor filter banks were used to analyze the directional response of the image and extract fiber orientation features. Fiber nodule regions in images are detected through morphological processing and Hough transform; the area, density, and distribution uniformity of the nodule regions are calculated as fibrous nodule features.

4. The control method for a hemp fiber combing machine based on digital analysis according to claim 3, characterized in that, When performing adaptive filtering and normalization processing on the fiber tension data to extract fiber strength features, the process includes: The fiber tension data is filtered using the minimum mean square error algorithm; the filtered fiber tension data is statistically analyzed to calculate the average tension value and tension fluctuation range; the initial fiber strength characteristics are determined based on the average tension value and tension fluctuation range; the initial fiber strength characteristics are normalized to finally form the fiber strength characteristics.

5. The control method for a hemp fiber combing machine based on digital analysis according to claim 4, characterized in that, When predicting fiber combing quality indicators using the support vector machine algorithm, the following are included: The fiber length distribution characteristics, fiber orientation characteristics, fiber nodule characteristics, and fiber strength characteristics are combined to form a four-dimensional feature vector; the similarity between the four-dimensional feature vector and each support vector in the support vector machine is calculated based on the radial basis function kernel; each similarity is multiplied by the corresponding Lagrange multiplier and summed; a bias term is added to the summation result to obtain a preliminary predicted value; the preliminary predicted value is linearly transformed to map it to the physical range of fiber combing quality indicators; the predicted values ​​of fiber separation degree, fiber parallelism, and fiber damage rate are output.

6. The control method for a hemp fiber combing machine based on digital analysis according to claim 5, characterized in that, When calculating the deviation vector based on the predicted fiber combing quality index and the preset target quality index, the following are included: The preset target quality indicators include fiber separation target value, fiber parallelism target value, and fiber damage rate target value; The fiber separation deviation is obtained by subtracting the target fiber separation value from the predicted fiber separation value; the fiber parallelism deviation is obtained by subtracting the target fiber parallelism value from the predicted fiber parallelism value. The fiber damage rate deviation is obtained by subtracting the predicted fiber damage rate from the target fiber damage rate. The fiber separation deviation, fiber parallelism deviation, and fiber damage rate deviation are normalized; the normalized fiber separation deviation, fiber parallelism deviation, and fiber damage rate deviation are then organized into a deviation vector.

7. The control method for a hemp fiber combing machine based on digital analysis according to claim 6, characterized in that, When dynamically adjusting the equipment operating parameters based on the deviation vector using a fuzzy PID control algorithm, the following is included: The deviation vector is fuzzified to generate a corresponding fuzzy set; fuzzy inference is performed based on the fuzzy set using a preset fuzzy rule base to generate a fuzzy inference result; the output is aggregated using the minimum-maximum synthesis method based on the fuzzy inference result to generate a fuzzy output; and defuzzification is performed using the centroid method to generate precise PID parameter adjustment values; the basic PID parameters are updated based on the PID parameter adjustment values ​​to generate adjusted PID parameters; the control quantity is calculated based on the adjusted PID parameters to generate a control quantity calculation result; the control quantity calculation result is converted into needle plate movement speed adjustment values, combing angle adjustment values, and draw ratio adjustment values, and the adjustment instructions for the equipment operating parameters are output.

8. The control method for a hemp fiber combing machine based on digital analysis according to claim 7, characterized in that, When monitoring the quality parameters of combed hemp fibers in real time through online quality inspection, the following parameters are included: Image data from the outlet of the combing machine is collected to generate a fiber distribution image. Based on the fiber distribution image, the fiber spacing uniformity is calculated using an image processing algorithm to generate a fiber separation parameter. The fiber orientation consistency is calculated using a direction analysis algorithm to generate a fiber parallelism parameter. Data on the breakage points of hemp fibers at the outlet of the combing machine are collected, and the number of breaks per unit length is calculated to generate a fiber damage rate parameter. Based on the fiber separation parameter, fiber parallelism parameter, and fiber damage rate parameter, a set of tissue formation quality parameters is established.

9. The control method for a hemp fiber combing machine based on digital analysis according to claim 8, characterized in that, When adjusting equipment operating parameters based on digital analysis, the following steps are included: Based on the quality deviation data, the mapping relationship between the feature vector and the quality index is updated through an incremental learning algorithm, and the parameter compensation coefficient is calculated by combining the environmental temperature and humidity change data. Based on the parameter compensation coefficients, the correction values ​​for needle plate movement speed, combing angle, and draft ratio are generated; and based on the correction values ​​for needle plate movement speed, combing angle, and draft ratio, the corrected equipment operating parameters are output.

10. A control system for a hemp fiber combing machine based on digital analysis, used to apply the control method for a hemp fiber combing machine based on digital analysis as described in any one of claims 1-9, characterized in that, include: The acquisition unit is configured to acquire fiber tension data, fiber distribution image data and equipment operating parameters in real time during the hemp fiber combing process based on a high-precision tension sensor and image acquisition device, and construct a real-time acquisition dataset. The processing unit is configured to perform multi-scale wavelet transform processing on the fiber distribution image data to extract fiber length distribution features, fiber orientation features, and fiber nodule features; The fiber tension data is subjected to adaptive filtering and normalization to extract fiber strength features; The adjustment unit is configured to form a four-dimensional feature vector based on the combination of fiber length distribution characteristics, fiber orientation characteristics, fiber nodule characteristics and fiber strength characteristics, and predict fiber combing quality indicators through a support vector machine algorithm; calculate the deviation vector based on the predicted fiber combing quality indicators and the preset target quality indicators, and dynamically adjust the equipment operating parameters based on the deviation vector through a fuzzy PID control algorithm. The correction unit is configured to monitor the set of quality parameters of the combed hemp fibers in real time through online quality detection. The set of quality parameters includes fiber separation parameters, fiber parallelism parameters, and fiber damage rate parameters. The unit compares the set of quality parameters with the predicted quality indicators to generate quality deviation data and corrects the equipment operating parameters based on digital analysis.

Images