Cold-rolled strip steel plate shape closed-loop control method and system based on dynamic feedback

Abstract

The invention provides a cold-rolled strip steel plate shape closed-loop control method and system based on dynamic feedback, and belongs to the technical field of cold-rolled strip steel production.The method comprises the steps that a dynamic incidence matrix and a plate shape deviation signal are generated based on cold-rolled strip steel transverse measuring point data, processing the dynamic incidence matrix to obtain a space correlation defect feature vector of the cold-rolled strip steel, and processing the strip shape deviation signal to obtain a multi-dimensional defect feature vector; determining the strip shape defect type of the cold-rolled strip steel based on the multi-dimensional defect feature vector and the space correlation defect feature vector; and strip steel production control parameters are adjusted based on the strip shape defect type. The plate shape quality stability of the cold-rolled strip steel can be improved.

Description

Technical Field

[0001] This application belongs to the technical field of cold-rolled strip steel production. More specifically, it relates to a shape closed-loop control method and system for cold-rolled strip steel based on dynamic feedback. Background Art

[0002] In the process of cold-rolled strip steel production, shape control plays a decisive role in product quality. Most of the existing technologies are based on static models for production control and lack effective tracking of the dynamic changes in the production process. During actual production, the working conditions of cold-rolled strip steel are constantly changing, while the static model cannot be adjusted in real time, resulting in the disconnection between the control strategy and the actual situation, making it difficult to ensure the accuracy of shape control and leading to poor stability of shape quality. Summary of the Invention

[0003] The purpose of this application is to provide a shape closed-loop control method and system for cold-rolled strip steel based on dynamic feedback, so as to improve the stability of the shape quality of cold-rolled strip steel.

[0004] In the first aspect of the embodiments of this application, a shape closed-loop control method for cold-rolled strip steel based on dynamic feedback is provided, including: Generating a dynamic correlation matrix and a shape deviation signal respectively based on the lateral measurement point data of cold-rolled strip steel, processing the dynamic correlation matrix to obtain a spatial correlation defect feature vector of cold-rolled strip steel, and processing the shape deviation signal to obtain a multi-dimensional defect feature vector; Determining the shape defect type of cold-rolled strip steel based on the multi-dimensional defect feature vector and the spatial correlation defect feature vector; Adjusting the production control parameters of the strip steel based on the shape defect type.

[0005] In the second aspect of the embodiments of this application, a shape closed-loop control system for cold-rolled strip steel based on dynamic feedback is provided, including: A defect feature extraction module, configured to generate a dynamic correlation matrix and a shape deviation signal respectively based on the lateral measurement point data of cold-rolled strip steel, process the dynamic correlation matrix to obtain a spatial correlation defect feature vector of cold-rolled strip steel, and process the shape deviation signal to obtain a multi-dimensional defect feature vector; A defect classification module, configured to determine the shape defect type of cold-rolled strip steel based on the multi-dimensional defect feature vector and the spatial correlation defect feature vector; A parameter adjustment module, configured to adjust the production control parameters of the strip steel based on the shape defect type.

[0006] In a third aspect of the embodiments of the present application, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the steps of the above-mentioned closed-loop control method for cold-rolled strip steel shape based on dynamic feedback are implemented.

[0007] In a fourth aspect of the embodiments of the present application, a computer-readable storage medium is provided. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned closed-loop control method for cold-rolled strip steel shape based on dynamic feedback are implemented.

[0008] The beneficial effects of the closed-loop control method and system for cold-rolled strip steel shape based on dynamic feedback provided by the embodiments of the present application are as follows: By constructing a dynamic correlation matrix and an attention enhancement mechanism, the embodiments of the present application obtain a spatial correlation defect feature vector. At the same time, a multi-dimensional defect feature vector is constructed based on multi-modal analysis. The combination of the two can comprehensively and accurately capture the complex features of strip steel shape defects, avoid the one-sidedness of single feature analysis, and improve the accuracy of defect recognition. On this basis, according to the defect type output by defect feature analysis, the embodiments of the present application can targetedly adjust the strip steel production control parameters, improve the control efficiency, reduce ineffective operations, and effectively guarantee the shape quality of the strip steel. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0010] Figure 1 It is a schematic flow chart of a closed-loop control method for cold-rolled strip steel shape based on dynamic feedback provided by an embodiment of the present application; Figure 2 It is a structural block diagram of a closed-loop control system for cold-rolled strip steel shape based on dynamic feedback provided by an embodiment of the present application; Figure 3 It is a schematic block diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0011] In the following description, specific details such as specific system structures and technologies are presented for illustration rather than limitation in order to thoroughly understand the embodiments of the present application. However, those skilled in the art should clearly understand that the present application can also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid unnecessary details from interfering with the description of the present application.

[0012] To make the objectives, technical solutions, and advantages of the present application clearer, the following will be described through specific embodiments in conjunction with the accompanying drawings.

[0013] Please refer to Figure 1 , Figure 1 which is a schematic flowchart of a shape closed-loop control method for cold-rolled strip steel based on dynamic feedback provided by an embodiment of the present application. The method may include S101 to S103.

[0014] S101: Generate a dynamic correlation matrix and a shape deviation signal respectively based on the lateral measurement point data of the cold-rolled strip steel, process the dynamic correlation matrix to obtain a spatial correlation defect feature vector of the cold-rolled strip steel, and process the shape deviation signal to obtain a multi-dimensional defect feature vector.

[0015] In this embodiment, generating a dynamic correlation matrix and a shape deviation signal respectively based on the lateral measurement point data of the cold-rolled strip steel, processing the dynamic correlation matrix to obtain a spatial correlation defect feature vector of the cold-rolled strip steel, and processing the shape deviation signal to obtain a multi-dimensional defect feature vector includes: Construct a dynamic correlation matrix based on the lateral measurement point data of the cold-rolled strip steel.

[0016] Perform attention enhancement on the dynamic correlation matrix to obtain a graph attention network.

[0017] Obtain a spatial correlation defect feature vector of the cold-rolled strip steel based on the graph attention network.

[0018] Generate a shape deviation signal based on the lateral measurement point data of the cold-rolled strip steel, and decompose the shape deviation signal into multi-modal components through adaptive parameters.

[0019] Construct a multi-dimensional defect feature vector based on the defect energy ratio, defect characteristic frequency, and lateral distribution entropy of the multi-modal components.

[0020] In this embodiment, constructing a dynamic correlation matrix based on the lateral measurement point data of the cold-rolled strip steel and performing attention enhancement on the dynamic correlation matrix to obtain a graph attention network includes: Take the lateral measurement point data of the cold-rolled strip steel as nodes, and extract the features of the lateral measurement point data of the cold-rolled strip steel as node features.

[0021] Construct a dynamic correlation matrix based on all node features.

[0022] Calculate the temperature difference, tension gradient, and flatness deviation correlation between two adjacent nodes. Determine the edge weight between the two adjacent nodes based on the temperature difference, tension gradient, and flatness deviation correlation.

[0023] Add the edge weight between each two adjacent nodes to the dynamic association matrix to obtain a graph attention network with enhanced attention.

[0024] In this embodiment, the node features include flatness deviation value, tension value, temperature value, and rolling speed. The lateral measurement point data of cold-rolled strip refers to the data collected at multiple measurement points set along the width direction of the cold-rolled strip during the production process. The flatness deviation value is used to measure the degree of difference between the actual shape and the ideal shape of the strip. The tension value refers to the magnitude of the tensile force applied to the strip, and the uneven distribution of tension may cause flatness defects in the strip. The temperature value refers to the surface temperature of the strip, and the change in temperature will affect the physical properties of the strip and thus affect the flatness. The rolling speed refers to the running speed of the strip during production.

[0025] The dynamic association matrix is a matrix that reflects the mutual relationship between each lateral measurement point of the cold-rolled strip. It is constructed based on the node features, and the elements reflect the degree of association between the measurement points. There is an edge between each two nodes, and the edge weight is used to represent the tightness of the association between adjacent nodes. The temperature difference refers to the temperature difference between two adjacent measurement points, the tension gradient refers to the degree of change in tension between adjacent measurement points, and the flatness deviation correlation is used to measure the similarity of flatness deviation between adjacent measurement points.

[0026] The graph attention network is obtained by enhancing the attention of the dynamic association matrix, which can enable the model to focus on the nodes related to flatness defects. The spatial association defect feature vector is a vector data that comprehensively reflects the spatial association relationship between each lateral measurement point of the cold-rolled strip and the potential flatness defect features. The spatial association defect feature vector contains information related to defects at specific spatial positions and can represent the distribution status and severity of flatness defects in the width direction of the strip.

[0027] In this embodiment, the lateral measurement point data of the cold-rolled strip is used as nodes, and its features are used as node features. The edge weight is determined by calculating the temperature difference, tension gradient, and flatness deviation correlation between adjacent nodes, and a dynamic association matrix is constructed accordingly to describe the relationship between the measurement points. Attention is enhanced for this matrix, that is, different attentions are given to different node associations according to the edge weight, so as to obtain a graph attention network that can highlight key information.

[0028] Exemplarily, this embodiment can use devices such as laser rangefinders, tension sensors, and infrared thermometers to collect the flatness deviation value, tension value, and temperature value of the lateral measurement points of the cold-rolled strip, and at the same time obtain the rolling speed from the mill control system. The collected data is preprocessed and feature-extracted to obtain node features.

[0029] In this embodiment, for every two adjacent nodes, the temperature difference, tension gradient, and shape deviation correlation between them are calculated, and the edge weights are determined by synthesizing these factors. A dynamic association matrix is constructed based on all node features, and the calculated edge weights are added to the corresponding positions. This embodiment processes the dynamic association matrix and adjusts the attention of the model to the association of each node according to the edge weights, obtaining a graph attention network with enhanced attention.

[0030] In this embodiment, the shape deviation signal is decomposed into multimodal components through adaptive parameters, including: The shape deviation signal is decomposed into edge wave mode components, center wave mode components, and local warping mode components through adaptive parameters.

[0031] The adaptive parameters include the number of modes and the quadratic penalty factor. The quadratic penalty factor is used to control the bandwidth of each mode component.

[0032] In this embodiment, a shape deviation signal is generated based on the lateral measurement point data of cold-rolled strip steel, including: Shape deviation data is generated based on the lateral measurement point data of cold-rolled strip steel, and the shape deviation data is preprocessed. The preprocessed shape deviation data is arranged in chronological order to obtain the shape deviation signal.

[0033] In this embodiment, the shape deviation signal is a one-dimensional signal sequence formed by arranging the shape deviation values of the lateral measurement points in chronological order, which is used to characterize the dynamic changes of the shape defects of the strip steel during the rolling process. In this embodiment, the number of modes is set to 3, and the shape deviation signal is decomposed into 3 typical defect modes (edge wave, center wave, local warping), which respectively correspond to the characteristic deformations of the edge, middle, and local regions of the strip steel.

[0034] The quadratic penalty factor is used to constrain the bandwidth of each mode during the variational mode decomposition process. By adjusting the quadratic penalty factor, the edge wave can correspond to the low-frequency band (<10Hz), the center wave can correspond to the middle-frequency band (10 - 20Hz), and the local warping can correspond to the high-frequency band (>20Hz), realizing the frequency separation of different defect modes.

[0035] The edge wave mode component is a characteristic component of the periodic fluctuation in the edge region extracted after the variational mode decomposition of the shape deviation signal, corresponding to the wavy deformation defect that appears at the edge of the strip steel. Its physical essence is that the uneven tension distribution or abnormal cooling rate in the edge region causes the elongation rate of the strip steel edge to be greater than or less than that of the middle, forming an edge wavy undulation.

[0036] The center wave mode component is a characteristic component that reflects the symmetric periodic fluctuation in the middle region of the strip steel, manifested as the wavy deformation in the middle part of the strip steel, which is caused by uneven middle pressure or abnormal thermal crown of the rolling rolls during the rolling process. Its fluctuation is symmetric, that is, the two edges are relatively flat, and the middle region has obvious undulations.

[0037] The local warping mode component is a high-frequency characteristic component that reflects the sudden deformation of a single point or a small area of the strip steel, manifested as the protrusion or depression of a local area of the strip steel, and is caused by sudden factors such as uneven local materials, foreign object pressing in, or blockage of the cooling nozzle.

[0038] The parameters related to the multi-modal component can include defect characteristic frequency, defect energy ratio, transverse distribution entropy, and typical inducing factors.

[0039] Exemplarily, the center wave mode component can be manifested as the wavy deformation of the middle part of the strip steel, such as in the range of 50%-80% of the width. The defect characteristic frequency range of the center wave mode component can be in the middle frequency band of 10-20 Hz, with medium-wavelength defects, reflecting the symmetric stress distribution characteristics of the middle region. The higher the defect energy ratio corresponding to the center wave mode component, the more serious the center wave defect. When the defect energy ratio is too high, the bending roll force or roll crown needs to be adjusted to balance the rolling pressure in the middle. The transverse distribution entropy corresponding to the center wave mode component indicates the width ratio of the defect distribution covering the middle. When the transverse distribution entropy is greater than 0.6, it indicates that the defect distribution covers more than 50% of the width in the middle. When the transverse distribution entropy is close to ln(n), it is a completely symmetric distribution, which is a typical center wave characteristic. The typical inducing factors of the center wave mode component can include: the rolling force in the middle is higher than the corresponding value at the edge, manifested as pressure concentration. The temperature in the middle of the roll is higher than the corresponding value at the edge by more than the corresponding value, manifested as abnormal thermal crown.

[0040] In this embodiment, the shape deviation data of each measuring point in the transverse direction of the strip steel can be collected in real time by devices such as a laser rangefinder. After preprocessing the collected shape deviation data, such as denoising and normalization, it is arranged in chronological order to form a continuous shape deviation signal, which contains the time-domain information of the overall shape fluctuation of the strip steel during the rolling process.

[0041] In this embodiment, the variational mode decomposition algorithm is used. With the number of modes as the constraint condition, the bandwidth of each mode is controlled by the quadratic penalty factor, and the shape deviation signal is decomposed into 3 mode components with different frequency characteristics. Among them, the edge wave mode component corresponds to the low-frequency fluctuation at the edge of the strip steel, the center wave mode component corresponds to the medium-frequency symmetric fluctuation in the middle of the strip steel, and the local warping mode component corresponds to the high-frequency mutation of a single point or a small area of the strip steel.

[0042] Energy analysis, spectrum analysis, and distribution uniformity calculation are performed on each mode component to obtain the defect energy ratio, defect characteristic frequency, and transverse distribution entropy respectively. The three are combined to form a multi-dimensional defect characteristic vector.

[0043] S102: Determine the shape defect type of the cold-rolled strip steel based on the multi-dimensional defect characteristic vector and the spatially correlated defect characteristic vector.

[0044] In this embodiment, the types of shape defects may include edge wave defects, center wave defects, and local warping defects.

[0045] Exemplarily, a multi-dimensional feature vector (9-dimensional) and a spatial correlation feature vector (256-dimensional) are input into the defect classification model. The defect classification model can be a three-layer fully connected neural network. The model outputs the probabilities of multiple defect types such as edge waves, center waves, local warping, and mixed defects through Softmax. The defect classification model can be trained with historical data to learn the energy distribution, frequency characteristics, and spatial correlation patterns of defects, such as the correlation between the symmetric distribution characteristics of center waves and the thermal crown of the rolls. Drive the corresponding actuators to adjust the relevant strip production control parameters according to the priority of edge wave > center wave > local warping. For example, for edge wave defects, the roll tilt angle is preferentially adjusted to balance the edge pressure. For center wave defects, the bending roll force is preferentially adjusted to correct the middle roll gap. For local warping defects, the opening degree of the segmented cooling valve is preferentially adjusted to suppress deformation through local cooling.

[0046] S103: Adjust the strip production control parameters based on the type of shape defect.

[0047] In this embodiment, after adjusting the strip production control parameters, a closed-loop control method for the shape of cold-rolled strip based on dynamic feedback further includes: Calculate the characteristic change amount of the multi-modal component before adjusting the target strip data and the strip production control parameters. The target strip data is the data of the cold-rolled strip on the cold-rolled strip production line after adjustment.

[0048] If the characteristic change amount of the multi-modal component does not meet the target condition, compensate and adjust the strip production control parameters based on the deviation data between the characteristic change amount of the multi-modal component and the target condition. The characteristic change amount of the multi-modal component includes the change amount of the defect energy ratio. The target condition includes that the change amount of the defect energy ratio is greater than or equal to the defect energy ratio change threshold.

[0049] In this embodiment, the characteristic change amount of the multi-modal component further includes the characteristic frequency regression value and the lateral distribution entropy correction value. The target condition further includes that the characteristic frequency regression value reaches the characteristic frequency regression target range, and the lateral distribution entropy correction value reaches the lateral distribution entropy correction target.

[0050] The actuators for realizing the adjustment of the strip production control parameters may include a segmented cooling valve controller, a roll tilt servo motor, and a bending roll hydraulic cylinder. These actuators can be used to adjust the following parameters: the opening degree of the segmented cooling valve: used to dynamically adjust the cooling water flow rate for the area with excessively high local temperature. The roll tilt angle: can adjust the horizontal tilt angle of the roll based on the lateral distribution deviation of the shape. The bending roll force compensation amount: can correct the bending roll force setting value in real time according to the tension fluctuation.

[0051] In this embodiment, the energy ratio change threshold refers to the minimum amount of defect energy reduction required after adjustment. For example, the defect energy corresponding to edge waves must decrease by 10%. The characteristic frequency regression range refers to the process tolerance range within which the defect frequency must fall. The distribution entropy correction target refers to the standard for defect distribution uniformity.

[0052] The defect priority adjustment strategy prioritizes the impact of shape defects on strip quality and downstream processing, setting the following priority: edge ripples > mid-waves > local warpage. Edge defects directly impact strip shearing accuracy and subsequent stamping. Failure to promptly correct these defects can lead to edge cracking, thus giving them the highest priority. Mid-wave deformation affects strip flatness and coating uniformity, but its impact is more manageable than that of edge ripples. Local defects, which have a smaller impact and can be quickly corrected through local cooling, have the lowest priority.

[0053] For example, this embodiment introduces a fast-response dynamic feedback verification mechanism. Assume that a cold-rolled strip production line is producing a batch of low-carbon steel strips with a specification of 0.8 mm thickness and 1200 mm width. During the production process, the method of this embodiment is used to control the plate shape. Through the analysis of multi-dimensional defect feature vectors and spatial correlation defect feature vectors, it is determined that the current strip has a fusion defect of edge waves and middle waves. According to the defect priority adjustment strategy, the edge wave defect is processed first, and the roll inclination angle is adjusted. Then the middle wave defect is processed and the bending roll force is adjusted.

[0054] Within 3 seconds of completing the initial adjustment, the strip was retested using a shape meter. Calculations revealed that the defect energy percentage of the edge wave modal component had decreased from 35% before the adjustment to 28%, despite the target of a decrease below 20%. The threshold for change in defect energy percentage was not reached. The characteristic frequency of the mid-wave modal component changed from 18Hz before the adjustment to 16Hz, despite the target frequency range of 10-14Hz, and thus did not return to the target range. The lateral distribution entropy of the edge and mid-wave components also failed to meet the correction target.

[0055] Based on the aforementioned non-compliance, a secondary compensation algorithm was activated. Based on the deviation between the characteristic variation of the multimodal components and the target conditions, further adjustments were made to the roll tilt angle and bending force. For example, the roll tilt angle was appropriately increased to further reduce the energy contribution of the edge wave defect, while the bending force was fine-tuned to bring the characteristic frequency of the mid-wave modal component back into the target range.

[0056] Within 3 seconds of the next adjustment, the shape meter remeasured the strip's shape deviation. This time, the defect energy percentage of the edge wave modal component dropped to 18%, meeting the energy percentage change threshold. The characteristic frequency of the mid-wave modal component was adjusted to 13 Hz, returning to the target range. The transverse distribution entropy also reached the correction target. At this point, the strip shape quality was determined to meet the requirements, completing this closed-loop feedback verification.

[0057] Exemplarily, the adjustment of multiple strip steel production control parameters may include: Using a simplified finite element method, the steel strip is divided into regions with 50mm node spacing. This is used to construct a cooling heat conduction model. Input data includes the current strip temperature field, target temperature uniformity (requiring lateral temperature differences to be controlled within 5°C), and cooling water flow constraints. Model predictive control is used to output the segmented cooling valve openings for the next three control cycles. The optimization process compensates for the approximately 200ms heat transfer delay from water to the strip. A deep Q-network, developed using reinforcement learning, is used to pre-train a library of optimal cooling strategies tailored to the strip's cooling characteristics. In actual production, basic control solutions can be quickly deployed based on the steel grade.

[0058] Based on the distribution of lateral deviations in the board shape, a local weighted regression method is used to fit the deviation curve and calculate the optimal tilt angle. During the calculation process, edge measurement points are given greater weight to ensure that side wave defects are corrected first.

[0059] A control variable conflict matrix was constructed to analyze the interplay between segmented cooling, roll tilt angle, and roll bending force. A multi-objective particle swarm optimization algorithm was used to optimize the coordinated parameters of segmented cooling, tilt angle, and roll bending force. The optimization objectives were to minimize plate shape deviation, energy consumption, and equipment life loss. Dynamic weighting factors were set to balance the relationships between these objectives.

[0060] As can be seen from the above, this embodiment utilizes a graph attention network to accurately extract spatially correlated defect feature vectors by constructing a dynamic correlation matrix and enhancing attention. Simultaneously, a multidimensional defect feature vector is constructed based on the multimodal decomposition of the flatness deviation signal. Combining these two components into a classification model comprehensively captures multidimensional defect characteristics, including energy, frequency, and spatial distribution. This significantly improves the accuracy of defect identification and classification, reduces misidentifications and missed detections, and provides a reliable basis for subsequent precise adjustments.

[0061] On this basis, this embodiment deeply couples the defect priority adjustment strategy with feature analysis results. Based on this defect priority adjustment strategy, this embodiment specifically adjusts strip production control parameters based on different defect types, prioritizing defects with the greatest impact on product quality. This improves control efficiency, reduces ineffective adjustments, ensures strip shape quality, and reduces production losses.

[0062] Through the feedforward control formed by the above-mentioned feature analysis and hierarchical regulation in this embodiment, combined with the target strip data compensation mechanism after parameter adjustment, a complete dynamic feedback closed loop is finally constructed. In this embodiment, compensation adjustment is performed according to the target strip data after parameter adjustment. By setting the threshold of the change amount of the defect energy ratio, the regression target range of the characteristic frequency, and the correction target of the transverse distribution entropy, the closed-loop control of dynamic feedback is realized, which can timely respond to various changes in the production process, ensure the stability of the strip shape quality, and improve the quality consistency of cold-rolled strip products.

[0063] In an embodiment of the present application, a multi-dimensional defect feature vector is constructed based on the defect energy ratio, defect characteristic frequency, and transverse distribution entropy of multi-modal components, including: Perform marginal spectrum analysis on the multi-modal components, and calculate the defect energy ratio, defect characteristic frequency, and transverse distribution entropy of each modal component respectively.

[0064] Construct a multi-dimensional defect feature vector corresponding to each modal component based on the defect energy ratio, defect characteristic frequency, and transverse distribution entropy of each modal component.

[0065] The defect energy ratio is used to quantify the severity of the defect of each modal component. The defect characteristic frequency represents the periodic fluctuation characteristics of the defect, and the transverse distribution entropy is used to describe the distribution uniformity of the defect in the transverse direction of the strip.

[0066] In this embodiment, after decomposing the strip shape deviation signal into multi-modal components through adaptive parameters, a closed-loop control method for the strip shape of cold-rolled strip based on dynamic feedback further includes: Calculate the defect energy ratio of the multi-modal components.

[0067] Perform a fast Fourier transform on the multi-modal components to obtain the spectral amplitude distribution, and take the frequency with the largest amplitude in the spectral amplitude distribution as the defect characteristic frequency of the multi-modal components.

[0068] Calculate the defect energy ratio of the multi-modal components of multiple strip sub-regions, and calculate the transverse distribution entropy based on the defect energy ratio of the multi-modal components of multiple strip sub-regions.

[0069] In this embodiment, the time-domain energy of each modal component is calculated by the sum of squares as the defect energy. If the defect energy ratio of the edge wave modal defect in the edge region is high, it indicates that the edge deformation is severe, and the roll tilt angle needs to be adjusted preferentially to balance the edge pressure. The defect energy ratio is directly related to the defect visibility and the impact on downstream processing. For example, when the defect energy ratio of the edge wave modal component is high, shear edge cracking defects are likely to appear at the strip edge.

[0070] The time-domain signal is converted into the frequency domain using the Fast Fourier Transform, and the frequency with the largest amplitude corresponds to the dominant fluctuation period of the defect. For example, the middle wave defect is caused by the periodic change of the thermal crown in the middle of the roll, and its characteristic frequency is strongly correlated with the roll rotation frequency. Different defect types correspond to different frequency ranges, and edge waves, middle waves, and local warping can be quickly distinguished through the characteristic frequencies.

[0071] The strip is divided into sub-regions transversely, and the energy proportion of each region is calculated. The distribution uniformity is measured through the information entropy formula. The entropy value reflects the defect morphology. For example, the low entropy value of local warping indicates a single-point anomaly, and local regulation needs to be carried out through the segmented cooling valve to avoid defect diffusion.

[0072] Exemplarily, 100 - 200 strip measuring points are evenly arranged along the width direction of the strip at the exit of the cold rolling mill, and the sampling frequency is 100 Hz. The shape deviation value of each measuring point is obtained in real time, and the tension value of the tension sensor, the temperature value of the infrared thermometer, and the rolling speed of the PLC system are synchronously collected.

[0073] Median filtering is used to remove the high-frequency noise of the shape deviation value, and then the shape deviation value is mapped to the [0, 1] interval through min-max normalization. The preprocessed shape deviation values are arranged in the order of acquisition time to form a one-dimensional signal sequence. Through the synchronous clock of the PLC system, it is ensured that the time stamps of the shape deviation signal, tension, temperature, and rolling speed data are consistent, providing a basis for subsequent multi-source fusion analysis.

[0074] According to the common defect types of cold-rolled strip, the number of modes is set to 3, corresponding to edge wave, middle wave, and local warping respectively. According to the rolling process experience, the quadratic penalty factor is initially set to 200, which is applicable to strip with a conventional thickness, such as 1 - 3 mm, and can be dynamically adjusted through the defect recognition accuracy in the future. For example, the quadratic penalty factor is 300 to enhance the separation of high-frequency components.

[0075] After obtaining the three modal components, calculate the proportion of defect energy of each modal component to quantify the severity of the defect. Perform the Fast Fourier Transform on each modal component to obtain the spectral amplitude distribution. Take the frequency with the largest amplitude in the spectrum as the defect characteristic frequency. The strip is divided into multiple regions transversely, calculate the proportion of energy of each modal component in each region, and calculate the information entropy (transverse distribution entropy) according to the proportion of energy of each modal component in each region. The smaller the information entropy value, the more concentrated the defect, such as local warping. The larger the entropy value, the more uniform the distribution, such as middle wave. Concatenate the proportion of defect energy, defect characteristic frequency, and transverse distribution entropy of the 3 modal components in order to form a 1×9-dimensional multi-dimensional defect feature vector. The multi-dimensional defect feature vector comprehensively characterizes the defect type, severity, fluctuation period, and distribution characteristics of the strip, providing a basis for subsequent defect classification and priority regulation.

[0076] In this embodiment, by constructing a multi-dimensional defect feature vector, the strip defect types, severity levels, fluctuation periods, and distribution characteristics can be comprehensively characterized. This embodiment can accurately quantify defects, quickly distinguish different defect types, reflect the defect morphology, provide a reliable basis for defect classification and priority regulation, and also helps to timely and specifically adjust the rolling parameters, improve the strip quality, reduce the risk of downstream processing defects, and enhance the production efficiency and product quality stability.

[0077] In one embodiment of the present application, determining the shape defect type of cold-rolled strip based on the multi-dimensional defect feature vector and the spatially correlated defect feature vector includes: Obtaining the defect probability of the multi-modal components based on the multi-dimensional defect feature vector.

[0078] Determining the defect spatial distribution probability based on the spatially correlated defect feature vector.

[0079] Determining the shape defect type of the cold-rolled strip based on the defect spatial distribution probability and the defect probability of the multi-modal components.

[0080] In this embodiment, the defect probability of the multi-modal components refers to the possibility of defects occurring in each modal component calculated by a certain method based on the multi-dimensional defect feature vector. For example, when the defect energy ratio of the edge wave modal component is too high, the defect characteristic frequency exceeds the normal range, or the transverse distribution entropy is abnormal, the probability of defects occurring in the edge wave will increase accordingly. The defect spatial distribution probability refers to the distribution possibility of the shape defect on the strip space determined based on the spatially correlated defect feature vector. For example, when the node characteristics in certain regions are abnormal and the edge weights between adjacent nodes are large, it indicates that the spatial correlation in these regions is tight, and the probability of defects occurring in these regions will increase.

[0081] This embodiment uses the energy ratio (reflecting the defect severity), characteristic frequency (distinguishing defect types), and transverse distribution entropy (locating the defect region) of the multi-modal components to output the probability of each modal defect (edge wave / middle wave / local warping) through a classification model.

[0082] This embodiment calculates the distribution probability of defects in the edge, middle, and local regions based on the spatially correlated features such as the temperature difference, tension gradient, and shape deviation correlation between measurement points extracted by the graph attention network.

[0083] This embodiment determines the final defect type by weighted fusion or rule matching, combining the defect probability and the spatial distribution probability. For example, a high probability in the edge region + a high probability of the edge wave modal is determined as an edge wave defect.

[0084] Exemplarily, the classification model can use a three-layer fully connected neural network, with a 9-dimensional input layer, 64 / 32 hidden layers, and a 3-dimensional output layer. The Softmax outputs the probabilities of various modal defects. Training data: More than 100,000 labeled historical samples, including single defects (such as pure edge waves) and mixed defects (such as edge waves + middle waves). The model is optimized with the cross-entropy loss function. Three regions are defined: the edge (100 mm on each side), the middle (60% of the width in the middle), and the local (±50 mm at a single point). Calculate the sum of the attention weights of the nodes in each region, and after normalization, obtain the distribution probability. For example, when the sum of the weights in the edge region is the largest, it indicates a high probability of edge defects.

[0085] The determination rules for single defects can include: The high probability of edge wave features matches the high defect distribution in the edge region, confirming that the defects are concentrated in the edge. The high probability of middle wave features is consistent with the symmetric distribution features in the middle region, confirming that the defects are concentrated in the middle. The high probability of high-frequency mutation features matches the concentrated distribution in the single-point region, confirming that the defect is a local sudden deformation.

[0086] Exemplarily, when the probability of edge wave defects exceeds 60% and the spatial distribution probability of defects in the edge region exceeds 50%, it is determined as a single edge wave defect.

[0087] When the probability of middle wave defects exceeds 60% and the spatial distribution probability of defects in the middle region exceeds 50%, it is determined as a single middle wave defect.

[0088] When the probability of local warping defects exceeds 60% and the spatial distribution probability of defects in the single-point region exceeds 40%, it is determined as a single local warping defect.

[0089] The determination rules for fused defects can include: When the probabilities of both edge wave defects and middle wave defects exceed 40% and the sum of the defect spatial distribution probabilities in the edge region and the middle region exceeds 100%, it is determined as a mixed defect of edge waves and middle waves. At this time, further compare the energy proportion of the two defects. The defect with a higher energy proportion is the dominant defect. For example, if the energy proportion of edge waves is greater than that of middle waves, it is marked as "edge wave-dominated mixed defect".

[0090] When the probability of local warping defects exceeds 30% and the spatial distribution probability of defects in the single-point region exceeds 30%, regardless of the existence of other defects, it is marked as "mixed defect containing local warping". Because the high-frequency sudden characteristics of local warping need to be identified first to avoid ignoring local anomalies due to the interference of other defects.

[0091] If the sum of the probabilities of all defect types is less than 80%, the system automatically triggers the data retest process. The strip steel is sampled again by the shape meter, and the multi-dimensional features and spatial correlation features are recalculated to avoid misjudgment caused by noise interference or data anomalies.

[0092] In this embodiment, the defects and distribution probabilities are calculated through multi-dimensional and spatial correlation feature vectors, and the defect types are determined by combining reasonable rules. It can accurately identify the defect types from complex strip shape data, providing a clear decision-making basis for subsequent priority regulation and parameter adjustment, and significantly improving the recognition accuracy and control efficiency of cold-rolled strip shape defects.

[0093] In one embodiment of the present application, the strip production control parameters are adjusted based on the strip shape defect types, including: If the strip shape defect type is a single defect, the strip production control parameters are adjusted based on this strip shape defect type. The strip production control parameters include the opening degree of the segmented cooling valve, the roll tilt angle, and the bending roll force.

[0094] If the strip shape defect type is a combined defect, the strip production control parameters are adjusted according to the defect priority adjustment strategy. The combined defects include at least two of the edge wave defect, the middle wave defect, and the local warping defect.

[0095] In this embodiment, the defect priority adjustment strategy is that the parameter adjustment priority corresponding to the edge wave defect is greater than the parameter adjustment priority corresponding to the middle wave defect, and the parameter adjustment priority corresponding to the middle wave defect is greater than the parameter adjustment priority corresponding to the local warping defect. A single defect means that the strip only has one typical strip shape defect, including the edge wave defect, the middle wave defect, or the local warping defect. A combined defect means that two or three typical defects exist simultaneously, such as edge wave + middle wave, middle wave + local warping, edge wave + middle wave + local warping, and it needs to be processed step by step according to the priority.

[0096] Exemplarily, a single defect is directly mapped to a unique control parameter. For example, the roll tilt angle is adjusted corresponding to the edge wave. The parameter adjustment amount is calculated through defect characteristics such as the defect energy ratio and the transverse distribution entropy.

[0097] The combined defects are adjusted corresponding to the strip production control parameters in the order of priority. In this embodiment, the high-priority defects are processed first. After their energy ratio drops below the threshold, the sub-priority defects are processed, avoiding multi-parameter coupling interference. For example, the adjustment of the tilt angle will indirectly affect the middle tension. The actuator actions are decoupled, and the physical action areas of the actuators corresponding to different defects are independent. The mechanical action conflicts are avoided by adjusting the order and time interval. For example, if the roll tilt and the bending roll force are adjusted simultaneously, it will cause the vibration of the roll system. In this embodiment, the single adjustment amplitude limit (such as the tilt angle ≤ 0.2° / time) is set to prevent new defects caused by excessive adjustment. For the secondary defects in the combined defects, the adjustment amplitude is executed at 50% of the dominant defect. For example, when there are edge wave + middle wave, the adjustment amplitude of the middle wave does not exceed half of the edge wave.

[0098] In this embodiment, for a single defect, the control parameters can be accurately matched and the adjustment amount can be determined according to the characteristics, so as to achieve efficient repair. In the face of a combined defect, it is processed step by step according to the priority strategy, which not only avoids the interference of parameter coupling, but also prevents mechanical conflicts and the generation of new defects by decoupling the actuator actions and limiting the adjustment amplitude. This greatly improves the flatness control accuracy, ensures the strip quality, improves the production stability and efficiency, and reduces the production cost.

[0099] A closed-loop flatness control method for cold-rolled strip based on dynamic feedback corresponding to the above embodiment Figure 2 is a structural block diagram of a closed-loop flatness control system for cold-rolled strip based on dynamic feedback provided in an embodiment of the present application. For the convenience of description, only the parts related to the embodiments of the present application are shown. Refer to Figure 2 This closed-loop flatness control system 20 for cold-rolled strip based on dynamic feedback includes: a defect feature extraction module 21, a defect classification module 22, and a parameter adjustment module 23.

[0100] Among them, the defect feature extraction module 21 is used to generate a dynamic correlation matrix and a flatness deviation signal respectively based on the lateral measurement point data of the cold-rolled strip, process the dynamic correlation matrix to obtain the spatial correlation defect feature vector of the cold-rolled strip, and process the flatness deviation signal to obtain a multi-dimensional defect feature vector.

[0101] The defect classification module 22 is used to determine the flatness defect type of the cold-rolled strip based on the multi-dimensional defect feature vector and the spatial correlation defect feature vector.

[0102] The parameter adjustment module 23 is used to adjust the strip production control parameters based on the flatness defect type.

[0103] In an embodiment of the present application, the defect feature extraction module 21 is specifically used to construct a dynamic correlation matrix based on the lateral measurement point data of the cold-rolled strip.

[0104] Enhance the attention of the dynamic correlation matrix to obtain a graph attention network.

[0105] Obtain the spatial correlation defect feature vector of the cold-rolled strip based on the graph attention network.

[0106] Generate a flatness deviation signal based on the lateral measurement point data of the cold-rolled strip, and decompose the flatness deviation signal into multi-modal components through adaptive parameters.

[0107] Construct a multi-dimensional defect feature vector based on the defect energy ratio, defect characteristic frequency, and lateral distribution entropy of the multi-modal components.

[0108] In an embodiment of the present application, the defect classification module 22 is specifically used to decompose the flatness deviation signal into an edge wave modal component, a center wave modal component, and a local warping modal component through adaptive parameters.

[0109] The adaptive parameters include the number of modes and the quadratic penalty factor. The quadratic penalty factor is used to control the bandwidth of each mode component.

[0110] In one embodiment of the present application, the defect classification module 22 is specifically configured to perform marginal spectrum analysis on multi-mode components, and calculate the defect energy proportion, defect characteristic frequency, and transverse distribution entropy of each mode component respectively.

[0111] Construct a multi-dimensional defect feature vector corresponding to each mode component based on the defect energy proportion, defect characteristic frequency, and transverse distribution entropy of each mode component.

[0112] The defect energy proportion is used to quantify the severity of defects in each mode component, the defect characteristic frequency represents the periodic fluctuation characteristics of defects, and the transverse distribution entropy is used to describe the distribution uniformity of defects in the transverse direction of the strip steel.

[0113] In one embodiment of the present application, the defect feature extraction module 21 is specifically configured to use the lateral measurement point data of the cold-rolled strip steel as nodes, and extract the features of the lateral measurement point data of the cold-rolled strip steel as node features. Construct a dynamic correlation matrix based on all node features.

[0114] Calculate the temperature difference, tension gradient, and shape deviation correlation between two adjacent nodes. Determine the edge weight between the two adjacent nodes based on the temperature difference, tension gradient, and shape deviation correlation.

[0115] Add the edge weight between each two adjacent nodes to the dynamic correlation matrix to obtain a graph attention network with enhanced attention.

[0116] In one embodiment of the present application, a shape closed-loop control system 20 for cold-rolled strip steel based on dynamic feedback further includes: a dynamic feedback adjustment module, configured to calculate the target strip steel data and the feature change amount of the multi-mode components before adjusting the strip steel production control parameters. The target strip steel data is the data of the cold-rolled strip steel on the cold-rolled strip steel production line after adjustment.

[0117] If the feature change amount of the multi-mode components does not meet the target condition, compensate and adjust the strip steel production control parameters based on the deviation data between the feature change amount of the multi-mode components and the target condition. The feature change amount of the multi-mode components includes the change amount of the defect energy proportion. The target condition includes that the change amount of the defect energy proportion is greater than or equal to the defect energy proportion change amount threshold.

[0118] In one embodiment of the present application, the parameter adjustment module 23 is specifically configured to, if the shape defect type is a single defect, adjust the strip steel production control parameters based on the shape defect type. The strip steel production control parameters include the opening degree of the sectional cooling valve, the roll tilt angle, and the bending roll force.

[0119] If the type of shape defect is a fusion defect, the strip production control parameters are adjusted according to the defect priority adjustment strategy. The fusion defect includes at least two of edge wave defect, middle wave defect, and local warping defect.

[0120] See Figure 3 , Figure 3 which is a schematic block diagram of an electronic device provided by an embodiment of the present application. As Figure 3 shown, the electronic device 300 in this embodiment may include: one or more processors 301, one or more input devices 302, one or more output devices 303, and one or more memories 304. The above-mentioned processors 301, input devices 302, output devices 303, and memories 304 communicate with each other through a communication bus 305. The memory 304 is used to store computer programs, and the computer programs include program instructions. The processor 301 is used to execute the program instructions stored in the memory 304. Among them, the processor 301 is configured to call the program instructions to execute the functions of each module in the above-mentioned system embodiments, such as Figure 2 the functions of the defect feature extraction module 21, defect classification module 22, and parameter adjustment module 23 shown.

[0121] It should be understood that in the embodiments of the present application, the so-called processor 301 may be a central processing unit (CPU), and this processor may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or this processor may also be any conventional processor, etc.

[0122] The input device 302 may include a touchpad, a fingerprint acquisition sensor (for acquiring the fingerprint information and fingerprint direction information of the user), a microphone, etc., and the output device 303 may include a display (such as an LCD), a speaker, etc.

[0123] The memory 304 may include a read-only memory and a random access memory, and provide instructions and data to the processor 301. A part of the memory 304 may also include a non-volatile random access memory. For example, the memory 304 may also store information about the device type.

[0124] In a specific implementation, the processor 301, input device 302, and output device 303 described in the embodiments of the present application may execute the implementation manners described in the first and second embodiments of a closed-loop control method for the shape of cold-rolled steel strips based on dynamic feedback provided by the embodiments of the present application, and may also execute the implementation manner of the electronic device 300 described in the embodiments of the present application, which will not be elaborated herein.

[0125] In another embodiment of the present application, a computer-readable storage medium is provided. The computer-readable storage medium stores a computer program, and the computer program includes program instructions. When the program instructions are executed by a processor, all or part of the processes in the method of the above embodiments are implemented. It can also be completed by instructing relevant hardware through the computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0126] The computer-readable storage medium can be the internal storage unit of the electronic device in any of the foregoing embodiments, such as the hard disk or memory of the electronic device. The computer-readable storage medium can also be an external storage device of the electronic device, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on the electronic device. Further, the computer-readable storage medium can also include both the internal storage unit and the external storage device of the electronic device. The computer-readable storage medium is used to store the computer program and other programs and data required by the electronic device. The computer-readable storage medium can also be used to temporarily store the data that has been output or will be output.

[0127] Those of ordinary skill in the art will realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0128] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described electronic devices and units can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0129] In several embodiments provided in this application, it should be understood that the disclosed electronic devices and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of units is only a logical function division, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed couplings or direct couplings or communication connections to each other can be indirect couplings or communication connections through some interfaces or units, or can be electrical, mechanical, or other forms of connection.

[0130] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiments of this application.

[0131] In addition, the functional units in each embodiment of this application can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0132] The above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art can easily think of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

Claims

1. A shape closed-loop control method for cold-rolled strip steel based on dynamic feedback, characterized in that Including: Generating a dynamic correlation matrix and a shape deviation signal respectively based on the lateral measurement point data of cold-rolled strip steel, processing the dynamic correlation matrix to obtain a spatial correlation defect feature vector of the cold-rolled strip steel, and processing the shape deviation signal to obtain a multi-dimensional defect feature vector; Determining the shape defect type of the cold-rolled strip steel based on the multi-dimensional defect feature vector and the spatial correlation defect feature vector; Adjusting the strip steel production control parameters based on the shape defect type.

2. The shape closed-loop control method for cold-rolled strip steel based on dynamic feedback according to claim 1, characterized in that The generating a dynamic correlation matrix and a shape deviation signal respectively based on the lateral measurement point data of cold-rolled strip steel, processing the dynamic correlation matrix to obtain a spatial correlation defect feature vector of the cold-rolled strip steel, and processing the shape deviation signal to obtain a multi-dimensional defect feature vector includes: Constructing a dynamic correlation matrix based on the lateral measurement point data of cold-rolled strip steel; Performing attention enhancement on the dynamic correlation matrix to obtain a graph attention network; Obtaining a spatial correlation defect feature vector of the cold-rolled strip steel based on the graph attention network; Generating a shape deviation signal based on the lateral measurement point data of cold-rolled strip steel, and decomposing the shape deviation signal into multi-modal components through adaptive parameters; Constructing a multi-dimensional defect feature vector based on the defect energy ratio, defect characteristic frequency, and lateral distribution entropy of the multi-modal components.

3. The shape closed-loop control method for cold-rolled strip steel based on dynamic feedback according to claim 2, wherein, The decomposing the shape deviation signal into multi-modal components through adaptive parameters includes: Decomposing the shape deviation signal into edge wave modal components, center wave modal components, and local warping modal components through adaptive parameters; The adaptive parameters include the number of modes and the quadratic penalty factor; the quadratic penalty factor is used to control the bandwidth of each modal component.

4. The shape closed-loop control method for cold-rolled strip steel based on dynamic feedback according to claim 2, characterized in that, The constructing a multi-dimensional defect feature vector based on the defect energy ratio, defect characteristic frequency, and lateral distribution entropy of the multi-modal components includes: Performing marginal spectrum analysis on the multi-modal components, and calculating the defect energy ratio, defect characteristic frequency, and lateral distribution entropy of each modal component respectively; Constructing a multi-dimensional defect feature vector corresponding to each modal component based on the defect energy ratio, the defect characteristic frequency, and the lateral distribution entropy of each modal component; The defect energy ratio is used to quantify the defect severity of each modal component, the defect characteristic frequency represents the periodic fluctuation characteristics of the defect, and the lateral distribution entropy is used to describe the distribution uniformity of the defect in the transverse direction of the strip steel.

5. A shape closed-loop control method for cold-rolled strip steel based on dynamic feedback according to claim 1, characterized in that The constructing a dynamic correlation matrix based on the lateral measurement point data of cold-rolled strip steel, and performing attention enhancement on the dynamic correlation matrix to obtain a graph attention network includes: Taking the lateral measurement point data of cold-rolled strip steel as nodes, and extracting the features of the lateral measurement point data of cold-rolled strip steel as node features; Constructing a dynamic correlation matrix based on all node features; Calculating the temperature difference, tension gradient, and shape deviation correlation between two adjacent nodes; determining the edge weight between the two adjacent nodes based on the temperature difference, tension gradient, and shape deviation correlation; Adding the edge weight between each two adjacent nodes to the dynamic correlation matrix to obtain a graph attention network after attention enhancement.

6. A shape closed-loop control method for cold-rolled strip steel based on dynamic feedback according to claim 1, characterized in that, Also including: Calculating the feature change amount of the target strip steel data and the multi-modal components before adjusting the strip steel production control parameters; The target strip data is the data of cold-rolled strip on the adjusted cold-rolled strip production line; If the feature change amount of the multimodal component does not meet the target condition, the strip production control parameters are compensated and adjusted based on the deviation data between the feature change amount of the multimodal component and the target condition; the feature change amount of the multimodal component includes the change amount of the defect energy ratio; the target condition includes that the change amount of the defect energy ratio is greater than or equal to the defect energy ratio threshold.

7. A shape closed-loop control method for cold-rolled strip steel based on dynamic feedback according to claim 1, characterized in that, The adjustment of the strip production control parameters based on the strip shape defect type includes: If the strip shape defect type is a single defect, the strip production control parameters are adjusted based on this strip shape defect type; the strip production control parameters include the opening degree of the segmented cooling valve, the roll tilt angle, and the bending roll force; If the strip shape defect type is a combined defect, the strip production control parameters are adjusted according to the defect priority adjustment strategy; the combined defect includes at least two of the edge wave defect, the middle wave defect, and the local warping defect.

8. A shape closed-loop control system for cold-rolled strip steel based on dynamic feedback, characterized in that, Including: A defect feature extraction module, configured to generate a dynamic correlation matrix and a strip shape deviation signal respectively based on the lateral measurement point data of the cold-rolled strip, process the dynamic correlation matrix to obtain a spatial correlation defect feature vector of the cold-rolled strip, and process the strip shape deviation signal to obtain a multi-dimensional defect feature vector; A defect classification module, configured to determine the strip shape defect type of the cold-rolled strip based on the multi-dimensional defect feature vector and the spatial correlation defect feature vector; A parameter adjustment module, configured to adjust the strip production control parameters based on the strip shape defect type.

9. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images