Method for controlling operation of sink roll for producing ultrathin strip steel

Abstract

The invention belongs to the technical field of sink roll production, and provides a method for controlling operation of a sink roll for producing ultrathin strip steel, which comprises the following steps of: acquiring dynamic tension-speed collaborative data flow of the ultrathin strip steel and three-dimensional vibration spectrum characteristics of the sink roll by deploying a multi-physics field coupling parameter monitoring system; constructing a time-frequency domain joint working condition feature library, adopting an improved support vector machine ensemble learning framework, and performing probability prediction on whether a current working condition is a transient working condition or not; and if the condition is predicted and judged to be a transient condition, key data features of the sink roll shaft are extracted, and a comprehensive risk assessment model is constructed based on a random forest model. In the monitoring stage, key sensor layout is focused, the transient working condition monitoring precision is improved, control lag caused by data redundancy is reduced, the belt breakage risk is reduced, nonlinear features of the transient working condition are accurately captured through time-frequency domain feature fusion and integrated learning, and the problem of transient fluctuation prediction lag is optimized.

Description

Technical Field

[0001] The present invention belongs to the technical field of submerged roll production, and specifically relates to a method for controlling the operation of submerged rolls for producing ultra-thin strip steel. Background Art

[0002] Ultra-thin strip steel, also known as extremely thin strip or ultra-thin steel strip, is a high-end category in steel materials, with a thickness usually between 0.05 mm and 0.2 mm (some high-end products can reach 0.015 mm, such as the "hand-tearable steel" of Taigang), and a width of more than 600 mm; The core role of the submerged roll in the production of ultra-thin strip steel is reflected in steering and friction control. It is completely driven by the frictional force between the strip steel and the roll surface. The surface grooves are designed to drain zinc liquid, ensure that the strip steel fits tightly, and prevent slipping or deviation; The passive drive characteristic of the submerged roll makes it extremely sensitive to tension fluctuations. Especially in transient working conditions, such as start-up and shutdown, speed mutation, or gauge change, tension fluctuations may cause the submerged roll to rotate unstably, exacerbating the risk of strip steel jitter or breakage. At the same time, tension fluctuations will be amplified in transient working conditions. Due to its small thickness and low rigidity, ultra-thin strip steel is more sensitive to tension changes and is more likely to deform or break; Therefore, the present invention provides a method for controlling the operation of submerged rolls for producing ultra-thin strip steel. Summary of the Invention

[0003] In order to make up for the deficiencies of the prior art and solve at least one of the technical problems proposed in the background art.

[0004] The technical solution adopted by the present invention to solve its technical problems is as follows: including the following steps: By deploying a multi-physical-field coupling parameter monitoring system, collecting the dynamic tension-velocity collaborative data stream of ultra-thin strip steel and the three-dimensional vibration spectrum characteristics of the submerged roll, constructing a time-frequency domain joint working condition feature library, and adopting an improved support vector machine integrated learning framework, probabilistically predict whether the current working condition is a transient working condition; If the prediction determines that it is a transient working condition, extract the key data characteristics of the submerged roll shaft, and based on the comprehensive risk assessment model constructed by the random forest model, divide the risk level of the transient working condition; The risk levels are: subcritical fluctuation, significant instability warning, production interruption critical; For the significant instability warning level among different risk levels, execute an optimization adjustment strategy.

[0005] As a further solution of the present invention: the deployment of the multi-physical-field coupling parameter monitoring system includes sensor layout; For the initial layout of the sensors, evaluate the importance of the sensors, and then adjust the positions of the important sensors based on the particle swarm optimization algorithm to obtain the optimized sensor layout.

[0006] As a further solution of the present invention: the important sensors are: obtain the important factors corresponding to the sensors; Extract the sensors whose important factors are greater than the important factor limit value as important sensors.

[0007] As a further solution of the present invention: the process of obtaining the important factors is: Quantify the importance of the sensors through the variance contribution rate and the spectral energy concentration; For any one sensor; count the number of times the importance is greater than the importance limit value, and calculate the ratio with the total number as the over-standard frequency ratio; for the case where the importance is greater than the importance limit value, calculate the difference between the importance and the importance limit value, and take the average value of the difference, and then calculate the ratio of the average value of the difference to the importance limit value as the over-standard amplitude ratio; add the over-standard frequency ratio and the over-standard amplitude ratio as the important factor.

[0008] As a further solution of the present invention: the process of obtaining the importance is: Calculate the ratio of the variance of each sensor data to the variance of all data as the variance contribution rate; perform spectral analysis on the vibration signal, and extract the proportion of the dominant frequency component in the total energy as the spectral energy concentration; add the variance contribution rate and the frequency energy concentration as the importance.

[0009] As a further solution of the present invention: the process of using the improved support vector machine ensemble learning framework to predict the probability of whether the current working condition is a transient working condition is: Use historical data to train the improved SVM ensemble learning framework, and the historical data includes: data samples of transient working conditions and non-transient working conditions; select the optimal model parameters through cross-validation; Preprocess and extract features from the sensor data collected in real time, obtain time-domain features, frequency-domain features, and time-frequency domain joint features and use them as inputs, input them into the trained working condition prediction model, and output the probability value that the working condition in the future time window is a transient working condition; If the output probability value is greater than the probability judgment threshold, it means that the working condition in the future time window is a transient working condition.

[0010] As a further solution of the present invention: the improved SVM ensemble learning framework is a machine learning framework that combines SVM and ensemble learning strategies; The ensemble learning strategies include Bagging ensemble, Boosting optimization, and Stacking fusion.

[0011] As a further solution of the present invention, the process of classifying the risk level of the transient condition is as follows: Input the extracted key data features into the comprehensive risk assessment model constructed based on the random forest model, and output the corresponding risk value; If the risk value is less than or equal to the first risk classification value, it is determined that the transient condition is a subcritical fluctuation risk level; If the risk value is greater than the first risk classification value and less than the second risk classification value, it is determined that the transient condition is a significant instability warning; If the risk value is greater than or equal to the second risk classification value, it is determined that the transient condition is critical for production interruption.

[0012] As a further solution of the present invention, the process of implementing the optimization adjustment strategy for the significant instability warning level in different risk levels is as follows: Obtain the importance features and the corresponding importance scores, and optimize and adjust the importance features in descending order of the importance scores; For each importance feature, use the ratio of its corresponding importance score to the total importance score of all importance features as the scaling factor; Calculate the deviation value between the current value and the target value of the importance feature, and multiply the deviation value by the scaling factor as the optimization adjustment amount.

[0013] As a further solution of the present invention, the process of obtaining the importance features and the importance scores is as follows: During the construction of each decision tree in the random forest, record the number of times each key data feature is used for node splitting; when a key data feature is used for node splitting in the decision tree, calculate the degree of node purity improvement brought by this split through the impurity reduction amount; For each key data feature, statistically calculate the total amount of impurity reduction when it is used as a split node in all decision trees, and then take the average value as the importance score of this feature; Extract the key data features corresponding to the importance score threshold greater than the importance score as the importance features.

[0014] The beneficial effects of the present invention are as follows: 1. In the monitoring stage, the present invention focuses on the layout of key sensors, improves the monitoring accuracy of transient conditions, reduces the control lag caused by data redundancy, reduces the risk of belt breakage, and accurately captures the non-linear features of transient conditions through time-frequency domain feature fusion and ensemble learning, optimizing the problem of prediction lag for transient fluctuations; 2. The present invention quantifies the impact degree of transient conditions on production, realizes the classification of risk levels, reduces misjudgment or missed judgment; identifies the features that contribute greatly to risk assessment, optimizes the problem that traditional methods cannot specifically evaluate risks, and provides data support for subsequent strategies; 3. The present invention implements precise measures for different risk levels, reduces unnecessary shutdowns, and improves production stability; dynamically adjusts parameters based on feature importance, optimizes the low efficiency of the one-size-fits-all adjustment method, optimizes the tension and speed stability of ultra-thin strip steel under transient conditions, and reduces the risk of strip breakage. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The present invention will be further described below in conjunction with the accompanying drawings.

[0016] Figure 1 is a flowchart of the steps of a method for controlling the operation of a submerged roll for producing ultra-thin strip steel according to the present invention; Figure 2 is a logic judgment diagram for evaluating the importance of sensors in a method for controlling the operation of a submerged roll for producing ultra-thin strip steel according to the present invention; Figure 3 is an architecture diagram of a system for controlling the operation of a submerged roll for producing ultra-thin strip steel according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] In order to make the technical means, creative features, achieved purposes and effects of the present invention easy to understand, the present invention will be further described below in conjunction with specific embodiments.

[0018] Example 1: During the production of ultra-thin strip steel, for the operation of the submerged roll, it is first necessary to predict the transient working conditions in advance, so that early warnings can be issued in a timely manner, and optimization adjustment strategies can be executed, thereby reducing the possible impact of transient working conditions on production stability; Please refer to Figure 1 and Figure 3 As shown, a method for controlling the operation of a submerged roll for producing ultra-thin strip steel according to an embodiment of the present invention includes the following steps: Step S10: By deploying a multi-physical-field coupling parameter monitoring system, collect the dynamic tension-velocity collaborative data stream of ultra-thin strip steel and the three-dimensional vibration frequency spectrum characteristics of the submerged roll, construct a time-frequency domain joint working condition feature library, and use an improved support vector machine integrated learning framework to perform probability prediction on whether the current working condition is a transient working condition; In this step, the deployment of the multi-physical-field coupling parameter monitoring system includes the selection of physical field sensors and the layout of sensors; Physical field sensors include mechanical field detection and thermal field detection. The mechanical field includes: a dynamic tension sensor using a piezoelectric tensiometer embedded in the bearing seat of the submerged roller; a speed sensor using a non-contact laser velocimeter to monitor the strip entrance and exit speeds; a three-dimensional vibration sensor using a three-axis accelerometer installed at the end of the submerged roller shaft to measure X / Y / Z axis vibration; the thermal field includes: a platinum-rhodium thermocouple array immersed in the zinc pot at different depths (surface, middle, and bottom) to measure temperature; and an infrared thermal imager for non-contact monitoring of the strip surface temperature distribution. For the initial layout of sensors, the importance of sensors is first evaluated, and then the positions of sensors are adjusted based on the particle swarm optimization algorithm to obtain the optimized sensor layout; Furthermore, the importance of sensors is quantified by variance contribution rate and spectrum energy concentration, and constructed into the fitness function of the particle swarm algorithm. This drives the optimization of sensor layout towards more important areas, achieving a balance between sensor quantity and monitoring performance. The importance of sensors is assessed based on variance contribution and spectrum energy concentration. The variance contribution is calculated as the ratio of the variance of each sensor's data to the variance of the total data. A larger variance indicates more severe fluctuations in the parameter monitored by the sensor. Spectral analysis is performed on the vibration signal, and the proportion of the dominant frequency component in the total energy is extracted as the spectrum energy concentration. A more concentrated energy indicates a location is more sensitive to specific fault modes (such as sinking roller resonance). The variance contribution and frequency energy concentration are added together to determine the importance. The variance contribution rate quantifies the sensor's contribution to the system's global fluctuations. The larger the variance, the more dramatic the parameter changes in the area where the sensor is located. This can guide resource allocation, prioritizing the deployment of high-precision sensors in areas with large fluctuations to improve the sensitivity of anomaly detection. Spectral energy concentration uses spectrum analysis to locate frequency bands sensitive to specific modes (such as resonance). A high energy concentration indicates that the location can effectively capture the characteristic frequency of the fault. Since the importance evaluation of sensors is calculated based on historical data of multiple productions, multiple importances are calculated for each sensor. For any sensor; Count the number of times the importance is greater than the importance limit, and calculate the ratio with the total number as the exceeding frequency ratio; for the case where the importance is greater than the importance limit, calculate the difference between the importance and the importance limit, take the mean of the difference, and then calculate the ratio of the mean of the difference to the importance limit as the exceeding amplitude ratio; add the exceeding frequency ratio and the exceeding amplitude ratio as the importance factor; extract the sensors corresponding to the values greater than the importance factor limit as important sensors.

[0019] Among them, the over-standard frequency ratio reflects the importance and stability of the sensor in multiple data. Screening can eliminate accidentally important sensors and retain long-term key sensors; the over-standard amplitude ratio quantifies the significance of the sensor's importance, reduces the resource occupation of marginal sensors, and ensures that the optimized layout maximizes the response ability to transient working conditions; Based on the particle swarm optimization algorithm, adjust the positions of important sensors. First, encode the sensor position coordinates as the particle dimension, randomly generate a particle swarm within the feasible domain. Each particle represents a sensor layout scheme. For each particle, calculate its comprehensive fitness function value, adjust the particle velocity according to the individual optimal and population optimal positions, and adjust the particle position according to the updated velocity; repeat the process of calculating fitness, updating velocity and position until the maximum number of iterations (such as 90 times) is reached or the fitness value converges; when the fitness value changes less than the threshold for 10 consecutive iterations, stop the iteration; Among them, the larger the comprehensive utilization value, the stronger the monitoring ability of the corresponding layout scheme for transient working conditions; Exemplarily, the axial position of the sink roll: x ∈ [0, 1000] mm; the radial position of the sink roll: x ∈ [0, 50] mm; the particle dimension D: D = 2N, where N is the number of important sensors. If N = 6, then D = 12. Each particle contains 12 coordinate values; randomly generate a particle swarm containing 50 particles within the feasible domain. Each particle represents a sensor layout scheme. For each layout scheme, calculate the comprehensive fitness function value. The comprehensive fitness function is the sum of the variance contribution rate and the spectral energy concentration; The above optimization of the importance evaluation and position layout of the sensor has at least the following effects in controlling the operation process of the sink roll in the production of ultra-thin strip steel: through the evaluation and screening of importance, it is possible to focus on sensors that have a great impact on transient working conditions and improve the monitoring accuracy of key parameters. Position optimization can make the sensors layout in key areas such as the sink roll shafting and the strip steel path; it can reduce the control instruction lag caused by data redundancy and increase the risk of strip breakage; In this step, constructing a time-frequency domain joint working condition feature library includes: extracting time domain and frequency domain features respectively, and then performing time-frequency domain feature fusion; After filtering and preprocessing the collected data, align it based on the time stamp to ensure the synchronization of feature extraction; For the extraction of time domain features, it includes mechanical field time domain features and thermal field time domain features respectively; among them, the mechanical field time domain features include but are not limited to: tension volatility, speed change rate, vibration energy; the thermal field time domain features include but are not limited to: zinc liquid temperature gradient, ultra-thin strip steel surface temperature uniformity; Tension volatility: Calculate the difference between the maximum value and the minimum value of the tension data, and then calculate the ratio of the difference to the mean value of the tension data; Rate of change: Calculate the difference between the speed value at the next moment and the speed value at the previous moment, and then calculate the ratio with the time difference between the two adjacent moments; Vibration energy: the root mean square value of the vibration signal within the window length; Zinc liquid temperature gradient: the temperature difference between the zinc pot surface and the bottom; Strip surface temperature uniformity: standard deviation of the strip surface temperature; The extraction of frequency domain features includes mechanical field frequency domain features and thermal field frequency domain features. The mechanical field frequency domain features include but are not limited to: the dominant frequency components of the vibration signal; the thermal field frequency domain features include but are not limited to: the energy of the temperature fluctuation spectrum. Dominant frequency components of the vibration signal: Perform a fast Fourier transform (FFT) on the vibration signal to extract the spectrum amplitude and identify the dominant frequency components, such as the rotation frequency of the sinking roll and the resonance frequency of the strip; Temperature fluctuation spectrum energy: Perform fast Fourier transform (FFT) on the zinc liquid temperature signal to extract the spectrum energy of a specific frequency band (such as 0.1-1 Hz); Fuse and splice the time domain features with the frequency domain features to construct the time-frequency domain joint features; Store time domain features, frequency domain features, and time-frequency domain joint features, and associate features with transient operating condition labels to form a time-frequency domain joint operating condition feature library; In this step, an improved support vector machine ensemble learning framework is used to perform a probability prediction on whether the future working condition is a transient working condition. The core principle of support vector machines is to map data into a high-dimensional space using kernel functions and find the optimal hyperplane to maximize the classification margin. In transient operating condition prediction, SVM can effectively handle joint time-frequency domain features and nonlinear relationships. The improved support vector machine ensemble learning framework is an advanced machine learning framework that combines SVM with ensemble learning strategies. It is used to improve the accuracy and robustness of transient operating condition prediction in ultra-thin strip production. Among them, the construction process of the improved support vector machine ensemble learning framework is: Kernel function optimization: Select kernel functions, including but not limited to RBF and polynomial kernels; optimize kernel parameters (such as the γ value of the RBF kernel) through grid search and cross-validation to improve model fitting ability; For example, the vibration spectrum characteristics collected by the three-axis accelerometer at the shaft end of the submerged roller are mapped using the RBF kernel to capture the nonlinear components in the vibration signal; Ensemble learning strategies include: Bagging integration, Boosting optimization and Stacking fusion; Bagging Ensemble: Generate multiple SVM base learners, each trained based on Bootstrap sampled data to reduce the model variance. The outputs of the base learners are fused through a voting mechanism to enhance the prediction stability. Boosting Optimization: Adopt the AdaBoost algorithm to weight the misclassified samples and iteratively optimize the model to improve the recognition ability for difficult samples. Stacking Fusion: Deploy a multi-layer SVM model to form a hierarchical prediction structure. Exemplarily, 50 SVM base learners are generated for Bagging ensemble. Each learner is trained based on different Bootstrap sampled data, and the outputs of these learners are fused by the majority voting method, significantly enhancing the prediction stability. During the Boosting optimization process, the AdaBoost algorithm is adopted to weight the misclassified transient condition samples. After 10 rounds of iterative optimization, the model's recognition ability for difficult samples is improved by 20%. In the Stacking fusion, a two-layer SVM model is deployed. The bottom layer model processes time-domain features, and the top layer model fuses frequency-domain features and time-frequency domain joint features. Based on the improved support vector machine ensemble learning framework and historical data, train a working condition prediction model. The specific process is as follows: Use historical data to train the improved SVM ensemble learning framework. The historical data includes data samples of transient conditions and non-transient conditions. The data samples are: input features (time-domain features, frequency-domain features, and time-frequency domain joint features), output labels (transient condition 1 and non-transient condition 0). Select the optimal model parameters through cross-validation, including but not limited to: kernel function type, regularization parameter C. Preprocess and extract features from the real-time collected sensor data to obtain time-domain features, frequency-domain features, and time-frequency domain joint features, and use them as inputs to the trained working condition prediction model to output the probability value that the working condition in the future time window is a transient condition. Among them, the future time window is set according to the characteristics of process parameters (such as strip thickness, running speed). For example, it is set to 0.1s, 0.5s, etc. Compare the output probability value with the probability judgment threshold. The probability judgment threshold is used to judge whether it is a transient condition, and its setting is based on the historical data set (including transient condition labels), and the optimal threshold is found through grid search. For example, in the range of strip thickness from 0.2 to 0.5mm and speed from 80 to 120m / min, the optimal threshold is P > 0.65. If the output probability value is greater than the probability judgment threshold, it indicates that the working condition in the future time window is a transient condition. If the output probability value is less than or equal to the probability judgment threshold, it indicates that the working condition within the future time window is a non-transient working condition; Step S20: If the prediction determines that it is a transient working condition, extract the key data features of the sink roll shaft, and based on the comprehensive risk assessment model constructed by the random forest model, classify the risk level of the transient working condition; Among them, the risk levels are: subcritical fluctuation, significant instability warning, production interruption critical; The risk of the transient working condition refers to that during the production process of ultra-thin strip steel, when the operating state of the sink roll undergoes sudden changes or abnormalities (i.e., transient working conditions), the transient working conditions may cause fluctuations in parameters such as the tension and speed of the ultra-thin strip steel, affecting the stability of the production process, and even leading to production accidents such as strip breakage. The above classification of the risk level is also the classification of the impact degree of the transient working condition on production stability; In this step, if it is a transient working condition, the process of extracting the key data features of the sink roll and constructing a comprehensive risk assessment model through the random forest model algorithm is as follows: Extract the key data features of the sink roll under transient working conditions from the multi-physical field coupling parameter monitoring system, including but not limited to: the instantaneous speed and speed volatility of the sink roll, the sudden change rate of strip steel tension, the peak value of tension fluctuation, and the change rate of zinc liquid temperature gradient; Input the extracted key data features into the comprehensive risk assessment model constructed based on the random forest model, and output the corresponding risk value; Among them, the comprehensive risk assessment model is a model pre-trained and constructed by those skilled in the art according to the operating characteristics and technical requirements of the sink roll during the production process of ultra-thin strip steel. This model uses the key data features of the sink roll as input and the risk value as output; Compare the risk value with the first risk division value and the second risk division value respectively; If the risk value is less than or equal to the first risk division value, it is determined that the transient working condition is at the subcritical fluctuation risk level; If the risk value is greater than the first risk division value and less than the second risk division value, it is determined that the transient working condition is a significant instability warning; If the risk value is greater than or equal to the second risk division value, it is determined that the transient working condition is production interruption critical; Optionally, during the process of using the random forest model to evaluate the risk level, it also includes the evaluation of the importance of key data features. The importance of the features is quantified by the number of times they are used for splitting in the decision tree and the degree of improvement of node purity after splitting; Specifically, during the construction of each decision tree in the random forest, the number of times each key data feature is used for node splitting is recorded; the more times it is used for splitting, the more important the key data feature is in the decision-making process; when a key data feature is used for node splitting in a decision tree, the degree of improvement in node purity brought about by this splitting is calculated; the degree of improvement in node purity can be measured by the reduction in impurity, such as the reduction in Gini impurity or information entropy; the greater the average reduction in impurity brought about by a feature in all decision trees, the higher the contribution of this feature to the model prediction result; For each key data feature, the total reduction in impurity when it is used as a splitting node in all decision trees is statistically calculated, and then the average value is taken as the importance score of this feature; According to a pre-set importance score threshold, the key data features corresponding to those greater than the importance score threshold are extracted as important features; The above evaluation of the importance score for key data features can identify the key data features that have a great impact on the prediction result, so that the key data features can be more targeted for selection and optimization, and a more accurate evaluation can be carried out; Step S30: Execute different optimization strategies for different risk levels; In this step, for transient conditions at the subcritical fluctuation risk level, a strategy of continuous monitoring is executed; For transient conditions at the critical point of production interruption, a strategy of interrupting the production process is executed; For transient conditions at the significant instability warning level, an optimization adjustment strategy is executed to optimize the risk caused by transient conditions to production stability; Specifically, the process of executing the optimization adjustment strategy is as follows: Based on the important features identified above and obtaining the corresponding importance scores, the important features are optimized and adjusted in descending order of importance scores; For each important feature, the ratio of its corresponding importance score to the total importance score of all important features is used as a scaling factor; Quantify the risk contribution degree through feature importance, and then use the scaling factor to achieve the optimal allocation of adjustment resources; Calculate the deviation value between the current value and the target value (i.e., the preset threshold of the operating process) of the important feature, and multiply the deviation value by the scaling factor as the optimization adjustment amount; After determining the optimization adjustment amount, corresponding control measures can be taken to adjust the value of the feature. Exemplarily, for the instantaneous speed of the immersion roll, the output power of the drive motor can be adjusted to change the speed; for the strip tension, the set value of the tension control system can be adjusted or the position of the immersion roll can be adjusted to change the tension; The main technical solution of this embodiment is as follows: By deploying a multi-physical-field coupling parameter monitoring system, the strip tension-velocity data stream and the submerged roll vibration spectrum are collected. After the importance evaluation of sensors (based on variance contribution rate and spectral energy concentration) and the optimized layout using the particle swarm algorithm, a time-frequency domain joint working condition feature library is constructed. Then, an improved support vector machine integrated learning framework is used to predict whether the future working condition is transient. If the prediction is a transient working condition, key data features such as the instantaneous rotational speed of the submerged roll and the strip tension mutation rate are extracted and input into a comprehensive risk assessment model based on a random forest model. The risk value is calculated and compared with a threshold, and it is divided into three levels of risk: "subcritical fluctuation", "significant instability warning", and "production interruption critical". During the process, key influencing factors are screened through feature importance evaluation (such as the number of splits and the improvement of node purity). Corresponding strategies are implemented according to the risk level. For subcritical fluctuations, continuous monitoring is carried out; for production interruption critical, production is immediately interrupted; when there is a significant instability warning, the optimization adjustment amount is calculated by multiplying the deviation value proportionality factor from high to low according to the feature importance score, reducing the impact of transient working conditions on production. This embodiment has at least the following beneficial effects: In the monitoring stage, the focus is on the layout of key sensors, improving the monitoring accuracy of transient working conditions, reducing control lag caused by data redundancy, and reducing the risk of strip breakage; through time-frequency domain feature fusion and integrated learning, the non-linear features of transient working conditions are accurately captured, optimizing the problem of prediction lag for transient fluctuations; quantifying the impact degree of transient working conditions on production, realizing the classification of risk levels, and reducing misjudgment or missed judgment; identifying features that contribute greatly to risk assessment, optimizing the problem that traditional methods cannot specifically evaluate risks, and providing data support for subsequent strategies; implementing targeted strategies for different risk levels, reducing unnecessary shutdowns, and improving production stability; dynamically adjusting parameters based on feature importance, optimizing the low efficiency of the one-size-fits-all adjustment method, optimizing the tension and speed stability of ultra-thin strip steel under transient working conditions, and reducing the risk of strip breakage.

[0020] Embodiment 2: Based on the same inventive concept as a method for controlling the operation of a submerged roll for producing ultra-thin strip steel in the foregoing embodiment, as Figure 2 shown, the present application provides a system for controlling the operation of a submerged roll for producing ultra-thin strip steel, wherein the system specifically includes: Transient working condition prediction module: By deploying a multi-physical-field coupling parameter monitoring system, the dynamic tension-velocity collaborative data stream of ultra-thin strip steel and the three-dimensional vibration spectrum characteristics of the submerged roll are collected, a time-frequency domain joint working condition feature library is constructed, and an improved support vector machine integrated learning framework is used to perform a probability prediction on whether the current working condition is a transient working condition. By deploying a multi-physical-field coupling parameter monitoring system, collecting strip tension-velocity data streams and submerged roll vibration spectra, after sensor importance evaluation (based on variance contribution rate and spectral energy concentration) and optimized layout using the particle swarm algorithm, a joint time-frequency domain working condition feature library is constructed, and then an improved support vector machine integrated learning framework is used to predict whether the future working condition is transient; Risk analysis module: If the prediction determines it to be a transient working condition, key data features of the submerged roll shaft are extracted, and based on a comprehensive risk assessment model constructed by a random forest model, the risk level of the transient working condition is classified; Among them, the risk levels are: subcritical fluctuation, significant instability warning, production interruption critical; If the prediction is a transient working condition, key data features such as the instantaneous speed of the submerged roll and the strip tension mutation rate are extracted, input into the comprehensive risk assessment model based on the random forest model, calculate the risk value and compare it with the threshold, and classify it into three levels of risk: "subcritical fluctuation", "significant instability warning", "production interruption critical". During the process, key influencing factors are screened through feature importance evaluation; Hierarchical optimization module: Different optimization strategies are executed for different risk levels; For transient working conditions at the subcritical fluctuation risk level, a strategy of continuous monitoring is executed; for transient working conditions at the production interruption critical level, a strategy of interrupting the production process is executed; for transient working conditions at the significant instability warning level, an optimization adjustment strategy is executed to optimize the risk caused by the transient working condition to production stability.

[0021] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification only illustrates the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for controlling the operation of submerged rolls for producing ultra-thin strip steel, characterized in that: It includes the following steps: By deploying a multi-physical-field coupling parameter monitoring system, collect the dynamic tension-velocity collaborative data stream of the ultra-thin strip steel and the three-dimensional vibration spectrum characteristics of the submerged roll, construct a time-frequency domain joint working condition feature library, and use an improved support vector machine integrated learning framework to probabilistically predict whether the current working condition is a transient working condition; If the prediction determines that it is a transient working condition, extract the key data characteristics of the submerged roll shaft, and based on the comprehensive risk assessment model constructed by the random forest model, divide the risk level of the transient working condition; The risk levels are: subcritical fluctuation, significant instability warning, production interruption critical; For the significant instability warning level among different risk levels, execute the optimization adjustment strategy.

2. A method for controlling the operation of a submerged roll for producing ultra-thin strip steel according to claim 1, characterized in that: The deployment of the multi-physical-field coupling parameter monitoring system includes sensor layout; For the initial layout of the sensors, evaluate the importance of the sensors, and then adjust the positions of the important sensors based on the particle swarm optimization algorithm to obtain the optimized sensor layout.

3. A method for controlling the operation of a submerged roll for producing ultra-thin strip steel according to claim 2, characterized in that: The important sensors are: obtain the important factors corresponding to the sensors; Extract the sensors whose important factors are greater than the important factor limit as important sensors.

4. A method for controlling the operation of a submerged roll for producing ultra-thin strip steel according to claim 2, characterized in that: The process of obtaining the important factors is: Quantify the importance of the sensors through the variance contribution rate and the spectral energy concentration; For any one sensor; count the number of times the importance is greater than the importance limit, and calculate the ratio with the total number as the over-standard frequency ratio; for the case where the importance is greater than the importance limit, calculate the difference between the importance and the importance limit, and take the average of the differences, and then calculate the ratio of the average of the differences to the importance limit as the over-standard amplitude ratio; add the over-standard frequency ratio and the over-standard amplitude ratio as the important factor.

5. A method for controlling the operation of a submerged roll for producing ultra-thin strip steel according to claim 4, characterized in that: The process of obtaining the importance is: Calculate the ratio of the variance of each sensor data to the variance of all data as the variance contribution rate; perform spectral analysis on the vibration signal, and extract the proportion of the dominant frequency component in the total energy as the spectral energy concentration; add the variance contribution rate and the frequency energy concentration as the importance.

6. A method for controlling the operation of a submerged roll for producing ultra-thin strip steel according to claim 1, characterized in that: The process of using the improved support vector machine integrated learning framework to probabilistically predict whether the current working condition is a transient working condition is: Use historical data to train the improved SVM integrated learning framework, and the historical data includes: data samples of transient working conditions and non-transient working conditions; select the optimal model parameters through cross-validation; Preprocess and extract features from the real-time collected sensor data, obtain time-domain features, frequency-domain features, and time-frequency domain joint features and use them as inputs, input them into the trained working condition prediction model, and output the probability value that the working condition in the future time window is a transient working condition; ​ 7. A method for controlling the operation of a submerged roll for producing ultra-thin strip steel according to claim 1, characterized in that: ​ ​ 8. A method for controlling the operation of a submerged roll for producing ultra-thin strip steel according to claim 1, characterized in that: ​ ​ If the risk value is less than or equal to the first risk division value, it is determined that the transient condition is at the subcritical fluctuation risk level; If the risk value is greater than the first risk division value and less than the second risk division value, it is determined that the transient condition is a significant instability warning; If the risk value is greater than or equal to the second risk division value, it is determined that the transient condition is critical for production interruption.

9. A method for controlling the operation of a submerged roll for producing ultra-thin strip steel according to claim 1, characterized in that: The process of implementing the optimization and adjustment strategy for the significant instability warning level in different risk levels is as follows: Obtain the importance features and their corresponding importance scores, and optimize and adjust the importance features in descending order of the importance scores; For each importance feature, use the ratio of its corresponding importance score to the total importance score of all importance features as the proportionality factor; Calculate the deviation value between the current value and the target value of the importance feature, and multiply the deviation value by the proportionality factor as the optimization and adjustment amount.

10. A method for controlling the operation of a submerged roll for producing ultra-thin strip steel according to claim 9, characterized in that: The process of obtaining the importance features and importance scores is as follows: During the construction of each decision tree in the random forest, record the number of times each key data feature is used for node splitting; when a key data feature is used for node splitting in the decision tree, calculate the degree of node purity improvement brought by this splitting through the impurity reduction amount; For each key data feature, count the total amount of impurity reduction when it is used as a splitting node in all decision trees, and then take the average value as the importance score of this feature; Extract the key data features corresponding to the importance score threshold greater than the threshold as the importance features.

Images