Winding control method and system based on tension sensor monitoring

Abstract

The application discloses a winding control method and system based on tension sensor monitoring, and mainly relates to the technical field of road construction. The method comprises the following steps: arranging a high-speed camera at both ends of a winding roller of a target winding machine, and arranging a tension sensor at both ends of a guide roller of the target winding machine; obtaining a left film image sequence, a right film image sequence, a left tension monitoring signal sequence and a right tension monitoring signal sequence; obtaining a screening left film transverse offset feature and a screening right film transverse offset feature; determining a left interactive tension monitoring signal feature and a right interactive tension monitoring signal feature; obtaining a winding joint feature; and transmitting the obtained adjustment winding roller speed and adjustment roller shaft angle to a control unit of the target winding machine for winding control. The application solves the technical problem that the existing technology lacks joint analysis of multi-dimensional data in the winding process, and the winding control does not match the actual situation, thereby improving the accuracy of winding control.

Description

Technical Field

[0001] This invention relates to the field of road construction technology, and specifically to a winding control method and system based on tension sensor monitoring. Background Technology

[0002] In the production and processing of roll materials such as films, paper, and metal foils, the winding process is a crucial step in ensuring the quality of the finished product. Real-time monitoring and control of the tension and lateral position of the roll material during winding are typically required to prevent defects such as wrinkling, breakage, burrs, and misalignment. Currently, traditional single-point tension monitoring cannot reflect the differences in tension distribution during winding, easily leading to undetected localized tension anomalies at the edges. Furthermore, controlling and adjusting based on a single offset during winding can result in overlooking situations where wrinkles occur in the middle while the edge offset is within the required range, thus affecting the winding quality.

[0003] Existing technologies lack joint analysis of multi-dimensional data during the winding process, resulting in a low degree of alignment between winding control and actual conditions. Summary of the Invention

[0004] This invention provides a winding control method and system based on tension sensor monitoring, which addresses the technical problem in the prior art that the lack of joint analysis of multi-dimensional data during the winding process leads to a low degree of alignment between winding control and actual conditions.

[0005] In view of the above problems, the present invention provides a winding control method and system based on tension sensor monitoring.

[0006] A first aspect of the present invention provides a winding control method based on tension sensor monitoring. The method includes: deploying high-speed cameras at both ends of the winding roller of a target winding machine to obtain left and right high-speed camera images; and deploying tension sensors at both ends of the guide roller of the target winding machine to obtain left and right tension sensors; when the target winding machine starts film winding, simultaneously activating the left and right high-speed cameras and the left and right tension sensors to monitor winding, obtaining a left film image sequence, a right film image sequence, a left tension monitoring signal sequence, and a right tension monitoring signal sequence; traversing the left and right film image sequences to perform lateral offset optical flow analysis, and... The results of the lateral offset optical flow analysis are integrated and filtered to obtain the lateral offset characteristics of the left and right films. The left and right tension monitoring signal sequences are then subjected to time-guided interaction feature analysis to determine the left and right interactive tension monitoring signal characteristics. A joint dual inconsistency analysis is performed on the left and right film lateral offset characteristics, as well as the left and right interactive tension monitoring signal characteristics, to obtain the winding joint characteristics. Based on these winding joint characteristics, the winding roller speed and roller angle are analyzed, and the obtained adjusted winding roller speed and roller angle are transmitted to the control unit of the target winding machine for winding control.

[0007] A second aspect of the present invention provides a winding control system based on tension sensor monitoring. The system includes: a sensor deployment module for deploying high-speed cameras at both ends of the winding roller of a target winding machine to obtain a left high-speed camera and a right high-speed camera, and for deploying tension sensors at both ends of the guide roller of the target winding machine to obtain a left tension sensor and a right tension sensor; a winding monitoring module for simultaneously activating the left high-speed camera, the right high-speed camera, and the left and right tension sensors to perform winding monitoring when the target winding machine starts film winding, obtaining a left film image sequence, a right film image sequence, a left tension monitoring signal sequence, and a right tension monitoring signal sequence; and an integrated filtering module for performing lateral offset optical flow analysis on the left and right film image sequences, and for filtering lateral offset signals. The optical flow analysis results are integrated and filtered to obtain the lateral offset features of the left and right films. A time-guided interactive feature analysis module is used to perform time-series guided interactive feature analysis on the left and right tension monitoring signal sequences to determine the left and right interactive tension monitoring signal features. A winding joint feature acquisition module is used to perform joint dual inconsistency analysis on the left and right film lateral offset features, as well as the left and right interactive tension monitoring signal features, to obtain the winding joint features. A winding control module is used to analyze the winding roller speed and roller angle based on the winding joint features, and transmit the obtained adjusted winding roller speed and roller angle to the control unit of the target winding machine for winding control.

[0008] One or more technical solutions provided in this invention have at least the following technical effects or advantages:

[0009] This invention involves placing high-speed cameras at both ends of the winding roller of a target winding machine to obtain images of a left-side high-speed camera and a right-side high-speed camera, and placing tension sensors at both ends of the guide roller of the target winding machine to obtain images of a left-side tension sensor and a right-side tension sensor. When the target winding machine starts film winding, the left-side high-speed camera, the right-side high-speed camera, and the left-side and right-side tension sensors are simultaneously activated to monitor the winding process, obtaining left-side film image sequences, right-side film image sequences, left-side tension monitoring signal sequences, and right-side tension monitoring signal sequences. Lateral offset optical flow analysis is performed by traversing the left-side and right-side film image sequences, and the results of the lateral offset optical flow analysis are integrated and filtered. The method involves obtaining the lateral offset characteristics of the left and right films during screening; performing time-guided interaction feature analysis on the left and right tension monitoring signal sequences to determine the interactive tension monitoring signal characteristics of the left and right sides; conducting joint dual inconsistency analysis on the lateral offset characteristics of the left and right films, as well as the interactive tension monitoring signal characteristics of the left and right sides, to obtain the joint winding characteristics; and analyzing the winding roller speed and roller angle based on the joint winding characteristics, then transmitting the obtained adjusted winding roller speed and roller angle to the control unit of the target winding machine for winding control. This achieves the technical effect of in-depth analysis of lateral offset and tension distribution during the winding process, effectively identifying the actual winding conditions, and thus improving the accuracy of winding control. Attached Figure Description

[0010] Appendix Figure 1 This is a schematic diagram of the winding control method based on tension sensor monitoring provided in an embodiment of the present invention.

[0011] Appendix Figure 2 This is a schematic diagram of the winding control system based on tension sensor monitoring provided in an embodiment of the present invention.

[0012] The labels shown in the attached diagram:

[0013] Sensor deployment module 11, winding monitoring module 12, integrated screening module 13, leading interactive feature analysis module 14, winding joint feature acquisition module 15, winding control module 16. Detailed Implementation

[0014] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined by the appended claims. It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to these processes, methods, products, or devices.

[0015] Example 1, as shown in the appendix Figure 1 As shown, the present invention provides a winding control method based on tension sensor monitoring, wherein the method includes:

[0016] Step A100: Install high-speed cameras at both ends of the winding roll of the target winding machine to obtain the left high-speed camera and the right high-speed camera, and install tension sensors at both ends of the guide roll of the target winding machine to obtain the left tension sensor and the right tension sensor.

[0017] Step A200: When the target winding machine starts film winding, the left high-speed camera, the right high-speed camera, the left tension sensor, and the right tension sensor are simultaneously activated to monitor the winding, and the left film image sequence, the right film image sequence, the left tension monitoring signal sequence, and the right tension monitoring signal sequence are obtained.

[0018] In one embodiment, to capture the lateral displacement of the film on the take-up roller during the operation of the target take-up machine, high-speed cameras are installed at both ends of the take-up roller. The high-speed cameras capture the lateral displacement of the film's ends from their intended positions after it is wound onto the take-up roller. Since the film is supported and transported by the guide rollers of the target take-up machine during winding, tension sensors installed at both ends of the guide rollers can capture the tension experienced by the film's edges during winding.

[0019] Furthermore, when the target winding machine starts film winding—that is, the moment when the film is transferred from the unwinding end to the winding roller via the guide roller and gradually forms a roll—the left and right high-speed cameras, as well as the left and right tension sensors, are simultaneously activated. The left and right high-speed cameras continuously capture images of the film edge at a high frame rate, forming multiple frames arranged chronologically to obtain the left and right film image sequences. Simultaneously with the high-speed camera image acquisition, the left and right tension sensors collect signals regarding the tension experienced by the film edge as it passes the guide roller, also arranged chronologically to obtain the left and right tension monitoring signal sequences.

[0020] Preferably, the left and right film image sequences reflect the lateral offset of the film at both ends on the winding roller and the development of the lateral offset during the monitoring time, while the left and right tension monitoring signal sequences reflect the tension changes on both sides of the film during the winding process.

[0021] By monitoring the film winding process from both visual and mechanical perspectives, the system achieves the technical effect of providing data support for subsequent comprehensive control and analysis of the winding machine from multiple dimensions, as well as providing basic data for subsequent analysis of data from different dimensions.

[0022] Step A300: Traverse the left thin film image sequence and the right thin film image sequence to perform lateral offset optical flow analysis, and integrate and filter the lateral offset optical flow analysis results to obtain the lateral offset features of the left thin film and the lateral offset features of the right thin film.

[0023] Furthermore, lateral offset optical flow analysis is performed on the left and right thin film image sequences, and the results of the lateral offset optical flow analysis are integrated and filtered to obtain the lateral offset features of the left and right thin films. In this embodiment of the invention, step A300 further includes:

[0024] Dense optical flow analysis is performed between any two adjacent thin film images in the left thin film image sequence and the right thin film image sequence to determine the lateral offset component of each pixel in each thin film image, thereby obtaining the left thin film lateral offset component set sequence and the right thin film lateral offset component set sequence.

[0025] Lateral offset feature analysis was performed on the left and right film lateral offset component set sequences to obtain the left and right film lateral offset feature sequences.

[0026] Intra-sequence feature integration and screening were performed on the left and right film lateral shift feature sequences to determine which film lateral shift features to screen.

[0027] In one possible embodiment, to capture the lateral offset of the film at both ends during winding, dense optical flow is used to analyze the left and right film image sequences respectively, determining the motion vector of each pixel in each film image. This motion vector includes a lateral offset component and a longitudinal offset component. Since the longitudinal direction continuously changes during film winding, we only need to focus on the lateral offset component to determine the lateral offset during the film winding process.

[0028] Preferably, the first and second left-side thin-film images located at the first and second positions are extracted from the left-side thin-film image sequence. Gaussian blur noise reduction is performed on the first and second left-side thin-film images using a 5×5 Gaussian kernel. Then, the horizontal displacement difference of each pixel is calculated to obtain the set of horizontal offset components of the first left-side thin film. Based on the same principle, dense optical flow analysis is performed on any two subsequent adjacent left-side thin-film images to determine the horizontal offset component of each pixel in each left-side thin-film image, thus obtaining the set of horizontal offset components of the left-side thin film. Similarly, the sequence of horizontal offset components of the right-side thin film is obtained.

[0029] The mean and maximum values ​​of the left and right film lateral offset component sets are calculated by traversing the sets. Each film lateral offset component has a direction, resulting in the following sequences: the mean sequence of the left film lateral offset component, the mean sequence of the right film lateral offset component, the maximum value sequence of the left film lateral offset component, and the maximum value sequence of the right film lateral offset component. The mean sequence of the left film lateral offset component and the maximum value sequence of the left film lateral offset component are mapped and correlated to obtain the left film lateral offset feature. The mean sequence of the right film lateral offset component and the maximum value sequence of the right film lateral offset component are also mapped and correlated to obtain the right film lateral offset feature. The left and right film lateral offset features respectively reflect the lateral offset of the film at both ends of the take-up roller.

[0030] Furthermore, by integrating and filtering the lateral offset feature sequences of the left and right films within the sequence, the more common offset situations on the left and right sides during the monitoring time can be captured from the sequence, which facilitates data support for subsequent adjustments to the winding roll speed and roll angle.

[0031] Furthermore, intra-sequence feature integration and screening are performed on the left and right film lateral offset feature sequences to determine which to screen for and which to screen for. In this embodiment of the invention, step A300 further includes:

[0032] Based on the left and right film lateral offset feature sequences, a set of left and right film lateral offset feature points is constructed.

[0033] The first left-side fitted line is obtained by fitting a straight line to the set of lateral offset feature points of the left thin film using the least squares method.

[0034] According to the preset integration and filtering diffusion step size, the neighborhood of the left fitted line is constructed to obtain the first left fitted line neighborhood;

[0035] The neighborhood of the first left-side fitted line is diffused again according to the preset integration and filtering diffusion step size to obtain the neighborhood of the second left-side fitted line.

[0036] Determine whether the neighborhood density of the second left fitted line neighborhood is greater than or equal to the neighborhood density of the first left fitted line neighborhood. If so, continue to diffuse the neighborhood edge of the second left fitted line neighborhood according to the preset integration and filtering diffusion step size until the preset stopping condition is met, obtain the target left fitted line neighborhood, and calculate the mean value of the left film lateral offset feature in the target left fitted line neighborhood to obtain the filtered left film lateral offset feature.

[0037] A straight line is fitted to the set of right-side film lateral offset feature points, and the first right-side fitted straight line obtained by fitting is integrated and filtered to obtain the filtered right-side film lateral offset features.

[0038] Furthermore, the preset stopping condition is that if the neighborhood density obtained in this iteration is less than or equal to the neighborhood density obtained in the previous iteration when the number of iterations is less than the preset number of iterations, the iteration stops.

[0039] When the number of iterations exceeds the preset number of iterations, if the difference between the density values ​​of the neighborhood obtained from two adjacent iterations is less than or equal to the preset difference in the density values ​​of the neighborhood, the iteration stops.

[0040] In one embodiment, two coordinate systems are constructed, with time as the horizontal axis and the left and right film lateral offset features as the vertical axes, respectively. The left and right film lateral offset feature sequences are then filled into these two coordinate systems to construct the left and right film lateral offset feature point sets, respectively. These sets reflect the lateral offset fluctuations on the left and right sides of the film during the monitoring period, respectively.

[0041] Preferably, a first left-side fitted line is obtained by fitting the set of lateral offset feature points of the left-side film using the least squares method, thus determining a fitted line that conforms to the general distribution of the set of lateral offset feature points of the left-side film. Then, using the left-side fitted line as the starting line, according to a preset integration and filtering diffusion step size pre-constructed by those skilled in the art, left-side film lateral offset feature points whose distances are within the preset integration and filtering diffusion step size are added to the neighborhood, thus obtaining the neighborhood of the first left-side fitted line. Furthermore, the number of left-side film lateral offset feature points contained in the neighborhood of the first left-side fitted line is counted, and the ratio of the count result to twice the preset integration and filtering diffusion step size is used as the neighborhood density of the neighborhood of the second left-side fitted line.

[0042] To determine whether the neighborhood of the first left-side fitted line already includes a densely distributed region from the set of lateral offset feature points of the left-side film, the neighborhood edges of the first left-side fitted line neighborhood are diffused outwards again according to a preset integration and filtering diffusion step size to obtain the second left-side fitted line neighborhood. The number of lateral offset feature points of the left-side film contained within the second left-side fitted line neighborhood is then counted again, and the result is compared to four times the preset integration and filtering diffusion step size to obtain the neighborhood density of the second left-side fitted line neighborhood. Similarly, in subsequent calculations of neighborhood density, the result is compared to the vertical distance between the two corresponding neighborhood edges.

[0043] Furthermore, the neighborhood density of the second left-side fitted line neighborhood is compared with that of the first left-side fitted line neighborhood. When the neighborhood density of the second left-side fitted line neighborhood is greater than or equal to that of the first left-side fitted line neighborhood, it indicates that the second left-side fitted line neighborhood is more representative of the distribution of feature points within the set of feature points of the left-side film lateral offset. Therefore, the neighborhood edge diffusion of the second left-side fitted line neighborhood continues according to the preset integration and filtering diffusion step size until the preset stopping condition is met. The preset stopping condition is that when the number of iterations is less than the preset number of iterations set by those skilled in the art, if the neighborhood density obtained in this iteration is less than or equal to the neighborhood density obtained in the previous iteration, the iteration stops, and the left-side fitted line neighborhood obtained in the previous iteration is taken as the target left-side fitted line neighborhood. When the number of iterations exceeds the preset number of iterations, if the difference between the neighborhood density values ​​obtained from two adjacent iterations is less than or equal to the preset neighborhood density value difference, the iteration stops even if the neighborhood density value obtained in the current iteration is less than or equal to the neighborhood density value obtained in the previous iteration. The neighborhood corresponding to the larger neighborhood density value obtained in the current iteration and the previous iteration is taken as the target left-side fitted line neighborhood. The mean value of the left-side film lateral offset feature within the target left-side fitted line neighborhood is calculated, which means calculating the mean value of the film lateral offset component and the mean value within the neighborhood of the maximum value of the film lateral offset component. The selected left-side film lateral offset feature is obtained. The selected left-side film lateral offset feature reflects the lateral offset of the two ends of the film during the monitoring time.

[0044] Based on the same principle as obtaining the lateral offset features of the left-side thin film, the set of lateral offset feature points of the right-side thin film is fitted with a straight line using the least squares method. The fitted first right-side line is then integrated and filtered to obtain the lateral offset features of the right-side thin film. By constructing the fitted line using the least squares method, the overall trend is extracted. Then, by combining neighborhood diffusion iteration, effective features that conform to the trend are selected. This preserves the main regularity of the lateral offset of the thin film while eliminating short-term random fluctuations, achieving the technical effect of improving the stability and reliability of lateral offset feature analysis.

[0045] Step A400: Perform time-series leading interaction feature analysis on the left tension monitoring signal sequence and the right tension monitoring signal sequence to determine the characteristics of the left interactive tension monitoring signal and the characteristics of the right interactive tension monitoring signal;

[0046] Furthermore, time-series leading interaction feature analysis is performed on the left tension monitoring signal sequence and the right tension monitoring signal sequence to determine the characteristics of the left interactive tension monitoring signal and the characteristics of the right interactive tension monitoring signal. In this embodiment of the invention, step A400 further includes:

[0047] Multiple time-series iterative interactive multi-scale analysis branches of the time-series iterative interactive multi-scale analysis channel are invoked to perform time-series signal feature identification on the left tension monitoring signal sequence, thereby obtaining multiple left tension monitoring signal features.

[0048] The similarity between each left-side tension monitoring signal feature and other left-side tension monitoring signal features is calculated through iteration, and the calculation results are averaged to obtain the average cross-similarity of multiple left-side tension monitoring signal features.

[0049] The left tension monitoring signal feature corresponding to the maximum value among the average interaction similarity values ​​of the multiple left tension monitoring signal features is taken as the leading left tension monitoring signal feature;

[0050] Based on the left-side tension monitoring signal features, fine-grained guidance interaction is performed on the multiple left-side tension monitoring signal features to obtain multiple left-side interactive tension monitoring signal features;

[0051] Calculate the mean value of the multiple left-side interactive tension monitoring signal characteristics to determine the left-side interactive tension monitoring signal characteristics;

[0052] Multiple time-series iterative interactive multi-scale analysis branches of the time-series iterative interactive multi-scale analysis channel are invoked to identify time-series signal features of the right-side tension monitoring signal sequence, and time-series interactive feature analysis is performed to determine the right-side interactive tension monitoring signal features.

[0053] Furthermore, step A400 in this embodiment of the invention also includes:

[0054] Obtain the historical tension anomaly log set of the target winding machine within a historical time period;

[0055] Based on the anomaly type, the historical tension anomaly log set is divided into heterogeneous categories to obtain multiple partitioned historical tension anomaly log sets.

[0056] Based on the timestamp of each historical tension anomaly log, determine the interval between two adjacent historical tension anomaly logs in multiple sets of historical tension anomaly logs, and obtain multiple sets of historical anomaly interval durations.

[0057] The maximum and minimum values ​​in the multiple sets of historical abnormal interval durations are extracted by traversing the data, and the union of the extraction results is obtained to obtain a multi-scale set for feature analysis.

[0058] Based on the aforementioned feature analysis multi-scale set, multiple time-series iterative interactive multi-scale analysis branches of the time-series iterative interactive multi-scale analysis channel are constructed.

[0059] In one embodiment, in order to reliably identify the tension conditions experienced by the film edges during the winding process, a time-series leading interaction feature analysis is performed on the left and right tension monitoring signal sequences through multi-scale, iterative, and lead-follow adjustment methods. This allows for in-depth exploration of the potential implicit relationships between the feature analyses of the tension monitoring signal sequences at different scales, thereby obtaining left and right interactive tension monitoring signal features that are more consistent with the actual tension conditions.

[0060] Preferably, by analyzing the tension anomalies of the target winding machine over a historical period, a multi-scale feature analysis set that matches the actual tension anomaly capture process is obtained. Then, based on the multi-scale feature analysis set, a time-series iterative interactive multi-scale analysis branch is constructed to perform multi-scale analysis on the tension monitoring signal sequence.

[0061] In one embodiment, the historical tension anomaly log set comprises log data from alarms triggered by abnormal conditions such as sudden increases, decreases, or continuous fluctuations in tension within a historical period, such as three months. Each historical tension anomaly log includes a timestamp, anomaly type, and anomaly duration. Using the anomaly type as an index, historical tension anomaly logs belonging to the same anomaly type are grouped into a set, resulting in multiple partitioned historical tension anomaly log sets. Then, based on the timestamp of each historical tension anomaly log, each partitioned historical tension anomaly log set is arranged in chronological order, resulting in multiple partitioned historical tension anomaly log sequences. Finally, the interval between two adjacent historical tension anomaly logs in each of the multiple partitioned historical tension anomaly log sequences is calculated sequentially from front to back, resulting in the multiple partitioned historical anomaly interval duration sets.

[0062] Furthermore, in order to reliably capture each anomaly type, the maximum and minimum values ​​within each of the multiple sets of historical anomaly interval durations are extracted, and the union of the extraction results is calculated. This enables simultaneous analysis from both a larger and smaller scale, avoiding anomaly omissions and achieving the goal of obtaining the multi-scale set of feature analysis.

[0063] Preferably, multiple abnormal tension signal segments are extracted from the historical tension anomaly log set, and feature annotations are performed on these segments to obtain multiple abnormal tension signal segments and their features as a training data set. Based on the anomaly types corresponding to different feature analysis scales, corresponding training data subsets are extracted from the training data set. The convolutional neural network framework constructed based on the corresponding feature analysis scale is trained using these training data subsets. For example, when the analysis scale is 20h, the convolutional neural network framework consists of two temporal convolutional layers. The first layer has a 3×1 kernel size, used to capture local fluctuations of three adjacent data points, with 16 output channels. The second layer has a 5×1 kernel size, used to capture the correlation features of five adjacent data points, with 32 output channels. The activation function for both layers is ReLU. Furthermore, the framework has one pooling layer with a 2×1 kernel size and a stride of 1, used to retain key features and reduce dimensionality. Finally, it is connected to a fully connected layer. A convolutional neural network framework is subjected to multiple rounds of supervised training using a subset of training data until training converges, resulting in a trained temporal iterative interactive multi-scale analysis branch. The input data for this branch is a tension monitoring signal sequence, and the output data is tension monitoring signal features. Based on the same principle, multiple temporal iterative interactive multi-scale analysis branches are constructed using a feature analysis multi-scale set. These branches are then connected in parallel to obtain the temporal iterative interactive multi-scale analysis channel.

[0064] Preferably, multiple time-series iterative interactive multi-scale analysis branches of the time-series iterative interactive multi-scale analysis channel are used to perform time-series signal feature convolution recognition on the left tension monitoring signal sequence to obtain signal features at different scales, which are the multiple left tension monitoring signal features. Then, the cosine similarity function is used to calculate the similarity between each left tension monitoring signal feature and other left tension monitoring signal features, and the average of the calculation results is obtained to obtain the average interaction similarity of multiple left tension monitoring signal features. The average interaction similarity of each left tension monitoring signal feature reflects the representativeness of each left tension monitoring signal feature under multi-scale analysis. The left tension monitoring signal feature corresponding to the maximum value of the average interaction similarity of multiple left tension monitoring signal features is used as the leading left tension monitoring signal feature. The leading left tension monitoring signal feature is the signal feature that best represents the global tension situation under different scale analyses, and is used to perform fine-grained leading interaction on other left tension monitoring signal features to reduce noise and bias. After fine-grained leading interaction analysis, the multiple following left interactive tension monitoring signal features are obtained. The average value of multiple left-side interactive tension monitoring signal features is calculated to obtain the left-side interactive tension monitoring signal features.

[0065] Based on the same principle as obtaining the characteristics of the left interactive tension monitoring signal, multiple time-series iterative interactive multi-scale analysis branches are invoked to identify the time-series signal characteristics of the right tension monitoring signal sequence, and the characteristics of the right tension monitoring signal are determined to perform time-series leading interactive feature analysis, thereby determining the characteristics of the right interactive tension monitoring signal.

[0066] Thus, multi-scale feature recognition is achieved for the left and right tension monitoring signal sequences respectively, comprehensively capturing tension fluctuations. Furthermore, fine-grained guidance interaction is performed by selecting leading features of the left and right tension monitoring signals to reduce interference from noise and abnormal fluctuations, thereby significantly enhancing the accuracy of feature extraction.

[0067] Furthermore, based on the leading left-side tension monitoring signal features, fine-grained leading interaction is performed on the multiple left-side tension monitoring signal features to obtain multiple following left-side interactive tension monitoring signal features. In this embodiment of the invention, step A400 further includes:

[0068] Calculate the fine-grained sub-feature similarity between the left-side tension monitoring signal feature and the multiple left-side tension monitoring signal features respectively to obtain multiple fine-grained sub-feature similarity sets;

[0069] The multiple fine-grained sub-feature similarity sets are normalized respectively, and multiple leading interaction matrices are constructed based on the normalization results;

[0070] By using multiple guiding interaction matrices to guide and interact with the multiple left-side tension monitoring signal features, the multiple left-side interactive tension monitoring signal features are obtained.

[0071] In one possible embodiment, the cosine similarity function is used to calculate the fine-grained sub-feature similarity between the leading left-side tension monitoring signal features and the multiple left-side tension monitoring signal features, respectively, for different sub-features such as maximum tension value, minimum tension value, peak value, valley value, instantaneous rate of change, and tension fluctuation frequency, thus obtaining a set of multiple fine-grained sub-feature similarities. Then, the multiple fine-grained sub-feature similarity sets are normalized within the set using the softmax formula, and the processed results are filled into an initially empty matrix to obtain the multiple leading interaction matrices. These multiple leading interaction matrices reflect the interaction weights of different sub-features between the leading left-side tension monitoring signal features and the multiple left-side tension monitoring signal features. Furthermore, multiple sample leading interaction matrices and multiple sample left-side tension monitoring signal features are obtained as training data, and the framework constructed based on a graph convolutional network is trained under supervision until convergence, obtaining a trained leading interactor. The leading interactor is then used to analyze the multiple leading interaction matrices and the corresponding multiple left-side tension monitoring signal features to obtain the multiple following left-side interactive tension monitoring signal features.

[0072] Step A500: Perform a joint dual inconsistency analysis on the lateral offset features of the left and right films of the screening, the left interactive tension monitoring signal features, and the right interactive tension monitoring signal features to obtain the joint winding features;

[0073] Furthermore, a joint dual inconsistency analysis is performed on the screening left-side film lateral offset features and screening right-side film lateral offset features, as well as the left-side interactive tension monitoring signal features and right-side interactive tension monitoring signal features, to obtain the winding joint features. In this embodiment of the invention, step A500 further includes:

[0074] The consistency of the winding offset is verified for the lateral offset features of the left and right screening films. If the verification is successful, the lateral offset features of the left and right screening films are added to the winding joint feature.

[0075] An inconsistency analysis of tension distribution is performed on the left-side interactive tension monitoring signal features and the right-side interactive tension monitoring signal features. The results of the inconsistency in tension distribution, the left-side interactive tension monitoring signal features, and the right-side interactive tension monitoring signal features are then added to the winding joint features.

[0076] Furthermore, if the verification fails, the camera positioned above the target winding machine is invoked to capture images of the winding roller and obtain real-time winding images.

[0077] The real-time winding image is subjected to winding defect identification to obtain winding defect features, and the winding defect features, the lateral offset features of the left screening film, and the lateral offset features of the right screening film are added to the winding joint features.

[0078] In one embodiment, on the one hand, it is necessary to jointly verify the lateral offset characteristics of the left and right sides of the screened film. Under normal circumstances, the left and right lateral offsets of the film on the winding roller should have the same directional bias, but the numerical difference should be within a certain range. However, when wrinkles appear on the winding roller, there will be significant differences in the numerical values ​​or inconsistent directional biases between the two sides of the lateral offset. In this case, relying solely on the lateral offset characteristics of the film on both sides is insufficient for reliable control of the film winding process. New features need to be added, namely, using a camera positioned above the target winding machine to capture images of the winding roller, obtain real-time winding images, and identify any winding anomalies reflected in the real-time winding images. On the other hand, it is necessary to jointly analyze the characteristics of the left and right interactive tension monitoring signals to determine the tension distribution, thereby better adjusting the winding roller speed and roller angle. Optionally, the cosine similarity function is used to identify the similarity between the left-side interactive tension monitoring signal features and the right-side interactive tension monitoring signal features. The identification result is then used as the tension distribution inconsistency result and added together with the left-side interactive tension monitoring signal features and the right-side interactive tension monitoring signal features into the winding joint features.

[0079] Preferably, it is determined whether the offset directions of the left-side film lateral offset feature and the right-side film lateral offset feature are consistent. If they are consistent, the offset difference between the left-side film lateral offset feature and the right-side film lateral offset feature is calculated to see if it is less than the offset difference threshold preset by those skilled in the art. If it is, the verification is successful, and no other features need to be added for analysis. The left-side film lateral offset feature and the right-side film lateral offset feature are added to the winding joint feature.

[0080] When the offset directions of the left-side film lateral offset feature and the right-side film lateral offset feature are inconsistent, or when the offset directions of the left-side film lateral offset feature and the right-side film lateral offset feature are consistent, but the offset difference between the left-side film lateral offset feature and the right-side film lateral offset feature is greater than the offset difference threshold preset by those skilled in the art, the verification fails, and new features need to be added to assist subsequent winding control.

[0081] Preferably, a camera positioned above the target winding machine is used to capture real-time images of the winding roller, obtaining a real-time winding image. This winding image reflects the current state of the film on the winding roller. A pre-trained winding defect identifier is then used to identify winding defects in the real-time winding image, obtaining winding defect features. These features, along with the lateral offset features of the film on the left and right sides of the screen, are then added to the joint winding features for subsequent winding control analysis.

[0082] Multiple real-time winding images and winding defect features of multiple samples are acquired as training data. The framework based on the convolutional neural network is trained under supervision. During the training process, the framework parameters are updated and adjusted according to the output results to obtain the winding defect recognizer after training.

[0083] Step A600: Analyze the winding roller speed and roller angle based on the winding joint features, and transmit the obtained winding roller speed and roller angle adjustment to the control unit of the target winding machine for winding control.

[0084] In one possible embodiment, a preset winding control model is obtained, wherein the input of the preset winding control model is winding joint features, and the output is adjusting the winding roller speed and adjusting the roller shaft angle. Multiple sample winding joint features and corresponding multiple sample adjusting winding roller speeds and multiple sample adjusting roller shaft angles are obtained as model training data. The model training data is divided into a 3:2 ratio according to a preset ratio to obtain a training set and a validation set. The framework based on a feedforward neural network is trained using the training set, and the multiple sample winding joint features from the validation set are input into the framework to obtain multiple output adjusting winding roller speeds and multiple output adjusting roller shaft angles. It is determined whether the deviation between the multiple output adjusting winding roller speeds and multiple output adjusting roller shaft angles and the multiple sample adjusting winding roller speeds and multiple sample adjusting roller shaft angles is less than a preset deviation threshold. If so, the training is complete. Then, the adjusting winding roller speeds and adjusting roller shaft angles are transmitted to the control unit of the target winding machine to control and adjust the winding roller speeds and roller shaft angles of the target winding machine.

[0085] In summary, the embodiments of the present invention have at least the following technical effects:

[0086] 1. This invention achieves pixel-level dynamic capture of the lateral offset of the film by deploying dual high-speed cameras at both ends of the winding roller and combining them with lateral offset optical flow analysis. After integrating and filtering out interfering data, the filtered offset features can accurately reflect the film edge alignment status, effectively avoiding the offset misjudgment caused by traditional single-view monitoring. Furthermore, by using dual tension sensors at both ends of the guide roller in conjunction with time-series guided interactive feature analysis, the instantaneous fluctuations of the tension signal are extracted through multi-scale branches. Combined with the guided-follow interactive mechanism to enhance feature synergy, the invention achieves the technical effect of reliably sensing the film winding process and thus improving the accuracy of winding control.

[0087] 2. This invention achieves comprehensive optimization of winding quality through combined dual inconsistency analysis and closed-loop control. It coordinates the analysis of offset and tension characteristics to determine the target winding control parameters. By supplementing the joint winding characteristics through combined dual analysis, it can fully reflect the target winding state, thereby achieving the technical effect of improving the reliability of winding control.

[0088] Example 2, based on the same inventive concept as the tension sensor-based winding control method in the preceding examples, as shown in the appendix. Figure 2 As shown, this invention provides a winding control system based on tension sensor monitoring. The system and method embodiments of this invention are based on the same inventive concept. The system includes:

[0089] The sensor deployment module 11 is used to deploy high-speed cameras at both ends of the winding roll of the target winding machine to obtain the left high-speed camera and the right high-speed camera, and to deploy tension sensors at both ends of the guide roll of the target winding machine to obtain the left tension sensor and the right tension sensor.

[0090] The winding monitoring module 12 is used to simultaneously activate the left high-speed camera, the right high-speed camera, the left tension sensor, and the right tension sensor to perform winding monitoring when the target winding machine starts film winding, and to obtain the left film image sequence, the right film image sequence, the left tension monitoring signal sequence, and the right tension monitoring signal sequence.

[0091] The integrated filtering module 13 is used to traverse the left thin film image sequence and the right thin film image sequence to perform lateral offset optical flow analysis, and to integrate and filter the lateral offset optical flow analysis results to obtain the lateral offset features of the left thin film and the lateral offset features of the right thin film.

[0092] The interactive feature analysis module 14 is used to perform time-series interactive feature analysis on the left tension monitoring signal sequence and the right tension monitoring signal sequence to determine the characteristics of the left interactive tension monitoring signal and the characteristics of the right interactive tension monitoring signal.

[0093] The winding joint feature acquisition module 15 is used to perform a joint dual inconsistency analysis on the lateral offset features of the left and right screening films, the left interactive tension monitoring signal features, and the right interactive tension monitoring signal features to obtain the winding joint features.

[0094] The winding control module 16 is used to analyze the winding roller speed and roller shaft angle based on the winding joint features, and transmit the obtained winding roller speed and roller shaft angle adjustment to the control unit of the target winding machine for winding control.

[0095] Furthermore, the integrated filtering module 13 is used to perform the following steps:

[0096] Dense optical flow analysis is performed between any two adjacent thin film images in the left thin film image sequence and the right thin film image sequence to determine the lateral offset component of each pixel in each thin film image, thereby obtaining the left thin film lateral offset component set sequence and the right thin film lateral offset component set sequence.

[0097] Lateral offset feature analysis was performed on the left and right film lateral offset component set sequences to obtain the left and right film lateral offset feature sequences, respectively.

[0098] Intra-sequence feature integration and screening were performed on the left and right film lateral shift feature sequences to determine which film lateral shift features to screen.

[0099] Furthermore, the integrated filtering module 13 is used to perform the following steps:

[0100] Based on the left and right film lateral offset feature sequences, a set of left and right film lateral offset feature points is constructed.

[0101] The first left-side fitted line is obtained by fitting a straight line to the set of lateral offset feature points of the left thin film using the least squares method.

[0102] According to the preset integration and filtering diffusion step size, the neighborhood of the left fitted line is constructed to obtain the first left fitted line neighborhood;

[0103] The neighborhood of the first left-side fitted line is diffused again according to the preset integration and filtering diffusion step size to obtain the neighborhood of the second left-side fitted line.

[0104] Determine whether the neighborhood density of the second left fitted line neighborhood is greater than or equal to the neighborhood density of the first left fitted line neighborhood. If so, continue to diffuse the neighborhood edge of the second left fitted line neighborhood according to the preset integration and filtering diffusion step size until the preset stopping condition is met, obtain the target left fitted line neighborhood, and calculate the mean value of the left film lateral offset feature in the target left fitted line neighborhood to obtain the filtered left film lateral offset feature.

[0105] A straight line is fitted to the set of right-side film lateral offset feature points, and the first right-side fitted straight line obtained by fitting is integrated and filtered to obtain the filtered right-side film lateral offset features.

[0106] Furthermore, the preset stopping condition is that if the neighborhood density obtained in this iteration is less than or equal to the neighborhood density obtained in the previous iteration when the number of iterations is less than the preset number of iterations, the iteration stops.

[0107] When the number of iterations exceeds the preset number of iterations, if the difference between the density values ​​of the neighborhood obtained from two adjacent iterations is less than or equal to the preset difference in the density values ​​of the neighborhood, the iteration stops.

[0108] Furthermore, the guiding interaction feature analysis module 14 is used to perform the following steps:

[0109] Multiple time-series iterative interactive multi-scale analysis branches of the time-series iterative interactive multi-scale analysis channel are invoked to perform time-series signal feature identification on the left tension monitoring signal sequence, thereby obtaining multiple left tension monitoring signal features.

[0110] The similarity between each left-side tension monitoring signal feature and other left-side tension monitoring signal features is calculated through iteration, and the calculation results are averaged to obtain the average cross-similarity of multiple left-side tension monitoring signal features.

[0111] The left tension monitoring signal feature corresponding to the maximum value among the average interaction similarity values ​​of the multiple left tension monitoring signal features is taken as the leading left tension monitoring signal feature;

[0112] Based on the left-side tension monitoring signal features, fine-grained guidance interaction is performed on the multiple left-side tension monitoring signal features to obtain multiple left-side interactive tension monitoring signal features;

[0113] Calculate the mean value of the multiple left-side interactive tension monitoring signal characteristics to determine the left-side interactive tension monitoring signal characteristics;

[0114] Multiple time-series iterative interactive multi-scale analysis branches of the time-series iterative interactive multi-scale analysis channel are invoked to identify time-series signal features of the right-side tension monitoring signal sequence, and time-series interactive feature analysis is performed to determine the right-side interactive tension monitoring signal features.

[0115] Furthermore, the guiding interaction feature analysis module 14 is used to perform the following steps:

[0116] Obtain the historical tension anomaly log set of the target winding machine within a historical time period;

[0117] Based on the anomaly type, the historical tension anomaly log set is divided into heterogeneous categories to obtain multiple partitioned historical tension anomaly log sets.

[0118] Based on the timestamp of each historical tension anomaly log, determine the interval between two adjacent historical tension anomaly logs in multiple sets of historical tension anomaly logs, and obtain multiple sets of historical anomaly interval durations.

[0119] The maximum and minimum values ​​in the multiple sets of historical abnormal interval durations are extracted by traversing the data, and the union of the extraction results is obtained to obtain a multi-scale set for feature analysis.

[0120] Based on the aforementioned feature analysis multi-scale set, multiple time-series iterative interactive multi-scale analysis branches of the time-series iterative interactive multi-scale analysis channel are constructed.

[0121] Furthermore, the guiding interaction feature analysis module 14 is used to perform the following steps:

[0122] Calculate the fine-grained sub-feature similarity between the left-side tension monitoring signal feature and the multiple left-side tension monitoring signal features respectively to obtain multiple fine-grained sub-feature similarity sets;

[0123] The multiple fine-grained sub-feature similarity sets are normalized respectively, and multiple leading interaction matrices are constructed based on the normalization results;

[0124] By using multiple guiding interaction matrices to guide and interact with the multiple left-side tension monitoring signal features, the multiple left-side interactive tension monitoring signal features are obtained.

[0125] Furthermore, the winding joint feature acquisition module 15 is used to perform the following steps:

[0126] The consistency of the winding offset is verified for the lateral offset features of the left and right screening films. If the verification is successful, the lateral offset features of the left and right screening films are added to the winding joint feature.

[0127] An inconsistency analysis of tension distribution is performed on the left-side interactive tension monitoring signal features and the right-side interactive tension monitoring signal features. The results of the inconsistency in tension distribution, the left-side interactive tension monitoring signal features, and the right-side interactive tension monitoring signal features are then added to the winding joint features.

[0128] Furthermore, if the verification fails, the camera positioned above the target winding machine is invoked to capture images of the winding roller and obtain real-time winding images.

[0129] The real-time winding image is subjected to winding defect identification to obtain winding defect features, and the winding defect features, the lateral offset features of the left screening film, and the lateral offset features of the right screening film are added to the winding joint features.

[0130] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0131] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0132] This specification and accompanying drawings are merely illustrative examples of the invention and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its scope. Therefore, if such modifications and modifications fall within the scope of the invention and its equivalents, the invention is intended to include these modifications and modifications.

Claims

1. A winding control method based on tension sensor monitoring, characterized in that, The method includes: High-speed cameras are installed at both ends of the winding roll of the target winding machine to obtain the left and right high-speed camera data. Tension sensors are installed at both ends of the guide roll of the target winding machine to obtain the left and right tension sensor data. When the target winding machine starts film winding, the left high-speed camera, the right high-speed camera, the left tension sensor and the right tension sensor are simultaneously activated to monitor the winding, and obtain the left film image sequence, the right film image sequence, the left tension monitoring signal sequence and the right tension monitoring signal sequence. Lateral offset optical flow analysis is performed by traversing the left thin film image sequence and the right thin film image sequence, and the results of the lateral offset optical flow analysis are integrated and filtered to obtain the lateral offset features of the left thin film and the lateral offset features of the right thin film. Time-series-guided interaction feature analysis was performed on the left tension monitoring signal sequence and the right tension monitoring signal sequence to determine the characteristics of the left interactive tension monitoring signal and the right interactive tension monitoring signal. A joint dual inconsistency analysis was performed on the lateral offset features of the left and right films of the screening film, as well as the interactive tension monitoring signal features of the left and right sides, to obtain the joint winding features. Based on the aforementioned winding joint features, the winding roller speed and roller shaft angle are analyzed, and the obtained winding roller speed and roller shaft angle are transmitted to the control unit of the target winding machine for winding control. Specifically, intra-sequence feature integration and screening are performed on the left and right film lateral migration feature sequences to determine which features to screen for. This includes: Based on the left and right film lateral offset feature sequences, a set of left and right film lateral offset feature points is constructed. The first left-side fitted line is obtained by fitting a straight line to the set of lateral offset feature points of the left thin film using the least squares method. According to the preset integration and filtering diffusion step size, the neighborhood of the left fitted line is constructed to obtain the first left fitted line neighborhood; The neighborhood of the first left-side fitted line is diffused again according to the preset integration and filtering diffusion step size to obtain the neighborhood of the second left-side fitted line. Determine whether the neighborhood density of the second left fitted line neighborhood is greater than or equal to the neighborhood density of the first left fitted line neighborhood. If so, continue to diffuse the neighborhood edge of the second left fitted line neighborhood according to the preset integration and filtering diffusion step size until the preset stopping condition is met, obtain the target left fitted line neighborhood, and calculate the mean value of the left film lateral offset feature in the target left fitted line neighborhood to obtain the filtered left film lateral offset feature. A straight line is fitted to the set of right-side film lateral offset feature points, and the first right-side fitted straight line obtained by fitting is integrated and filtered to obtain the filtered right-side film lateral offset features.

2. The winding control method based on tension sensor monitoring as described in claim 1, characterized in that, Lateral offset optical flow analysis is performed by traversing the left and right thin film image sequences, and the results of the lateral offset optical flow analysis are integrated and filtered to obtain lateral offset features of the left and right thin films, including: Dense optical flow analysis is performed between any two adjacent thin film images in the left thin film image sequence and the right thin film image sequence to determine the lateral offset component of each pixel in each thin film image, thereby obtaining the left thin film lateral offset component set sequence and the right thin film lateral offset component set sequence. Lateral offset feature analysis was performed on the left and right film lateral offset component set sequences to obtain the left and right film lateral offset feature sequences. Intra-sequence feature integration and screening were performed on the left and right film lateral shift feature sequences to determine which film lateral shift features to screen.

3. The winding control method based on tension sensor monitoring as described in claim 2, characterized in that, The preset stopping condition is that if the neighborhood density obtained in the current iteration is less than or equal to the neighborhood density obtained in the previous iteration, the iteration stops when the number of iterations is less than the preset number of iterations. When the number of iterations exceeds the preset number of iterations, if the difference between the density values ​​of the neighborhood obtained from two adjacent iterations is less than or equal to the preset difference in the density values ​​of the neighborhood, the iteration stops.

4. The winding control method based on tension sensor monitoring as described in claim 1, characterized in that, A time-leading interaction feature analysis is performed on the left-side tension monitoring signal sequence and the right-side tension monitoring signal sequence to determine the characteristics of the left-side interactive tension monitoring signal and the characteristics of the right-side interactive tension monitoring signal, including: Multiple time-series iterative interactive multi-scale analysis branches of the time-series iterative interactive multi-scale analysis channel are invoked to perform time-series signal feature identification on the left tension monitoring signal sequence, thereby obtaining multiple left tension monitoring signal features. The similarity between each left-side tension monitoring signal feature and other left-side tension monitoring signal features is calculated through iteration, and the calculation results are averaged to obtain the average cross-similarity of multiple left-side tension monitoring signal features. The left tension monitoring signal feature corresponding to the maximum value among the average interaction similarity values ​​of the multiple left tension monitoring signal features is taken as the leading left tension monitoring signal feature; Based on the left-side tension monitoring signal features, fine-grained guidance interaction is performed on the multiple left-side tension monitoring signal features to obtain multiple left-side interactive tension monitoring signal features; Calculate the mean value of the multiple left-side interactive tension monitoring signal characteristics to determine the left-side interactive tension monitoring signal characteristics; Multiple time-series iterative interactive multi-scale analysis branches of the time-series iterative interactive multi-scale analysis channel are invoked to identify time-series signal features of the right-side tension monitoring signal sequence, and time-series interactive feature analysis is performed to determine the right-side interactive tension monitoring signal features.

5. The winding control method based on tension sensor monitoring as described in claim 4, characterized in that, include: Obtain the historical tension anomaly log set of the target winding machine within a historical time period; Based on the anomaly type, the historical tension anomaly log set is divided into heterogeneous categories to obtain multiple partitioned historical tension anomaly log sets. Based on the timestamp of each historical tension anomaly log, determine the interval between two adjacent historical tension anomaly logs in multiple sets of historical tension anomaly logs, and obtain multiple sets of historical anomaly interval durations. The maximum and minimum values ​​in the multiple sets of historical abnormal interval durations are extracted by traversing the data, and the union of the extraction results is obtained to obtain a multi-scale set for feature analysis. Based on the aforementioned feature analysis multi-scale set, multiple time-series iterative interactive multi-scale analysis branches of the time-series iterative interactive multi-scale analysis channel are constructed.

6. The winding control method based on tension sensor monitoring as described in claim 4, characterized in that, Based on the aforementioned left-side tension monitoring signal characteristics, fine-grained guidance interaction is performed on the multiple left-side tension monitoring signal characteristics to obtain multiple left-side interaction tension monitoring signal characteristics, including: Calculate the fine-grained sub-feature similarity between the left-side tension monitoring signal feature and the multiple left-side tension monitoring signal features respectively to obtain multiple fine-grained sub-feature similarity sets; The multiple fine-grained sub-feature similarity sets are normalized respectively, and multiple leading interaction matrices are constructed based on the normalization results; By using multiple guiding interaction matrices to guide and interact with the multiple left-side tension monitoring signal features, the multiple left-side interactive tension monitoring signal features are obtained.

7. The winding control method based on tension sensor monitoring as described in claim 1, characterized in that, A joint dual inconsistency analysis is performed on the lateral offset features of the left and right films, the left interactive tension monitoring signal features, and the right interactive tension monitoring signal features to obtain the joint winding features, including: The consistency of the winding offset is verified for the lateral offset features of the left and right screening films. If the verification is successful, the lateral offset features of the left and right screening films are added to the winding joint feature. An inconsistency analysis of tension distribution is performed on the left-side interactive tension monitoring signal features and the right-side interactive tension monitoring signal features. The results of the inconsistency in tension distribution, the left-side interactive tension monitoring signal features, and the right-side interactive tension monitoring signal features are then added to the winding joint features.

8. The winding control method based on tension sensor monitoring as described in claim 7, characterized in that, If the verification fails, the camera positioned above the target winding machine is invoked to capture images of the winding roller and obtain real-time winding images. The real-time winding image is subjected to winding defect identification to obtain winding defect features, and the winding defect features, the lateral offset features of the left screening film, and the lateral offset features of the right screening film are added to the winding joint features.

9. A winding control system based on tension sensor monitoring, characterized in that, The system is used to implement the winding control method based on tension sensor monitoring as described in any one of claims 1-8, the system comprising: The sensor deployment module is used to deploy high-speed cameras at both ends of the winding roll of the target winding machine to obtain the left high-speed camera and the right high-speed camera, and to deploy tension sensors at both ends of the guide roll of the target winding machine to obtain the left tension sensor and the right tension sensor. The winding monitoring module is used to simultaneously activate the left high-speed camera, the right high-speed camera, the left tension sensor, and the right tension sensor to monitor the winding when the target winding machine starts film winding, and to obtain the left film image sequence, the right film image sequence, the left tension monitoring signal sequence, and the right tension monitoring signal sequence. An integrated filtering module is used to traverse the left thin film image sequence and the right thin film image sequence to perform lateral offset optical flow analysis, and to integrate and filter the lateral offset optical flow analysis results to obtain the lateral offset features of the left thin film and the lateral offset features of the right thin film. The interactive feature analysis module is used to perform time-series interactive feature analysis on the left tension monitoring signal sequence and the right tension monitoring signal sequence to determine the characteristics of the left interactive tension monitoring signal and the characteristics of the right interactive tension monitoring signal. The winding joint feature acquisition module is used to perform a joint dual inconsistency analysis on the lateral offset features of the left and right screening films, the left interactive tension monitoring signal features, and the right interactive tension monitoring signal features to obtain the winding joint features. The winding control module is used to analyze the winding roller speed and roller shaft angle based on the winding joint features, and transmit the obtained winding roller speed and roller shaft angle adjustment to the control unit of the target winding machine for winding control.

Images