A full closed mode rolling method, system, device and storage medium

Abstract

The application relates to the technical field of intelligent manufacturing, in particular to a rolling method, system and device in a full-closed mode and a storage medium, wherein the method comprises the following steps: collecting plate thickness and temperature data in real time through a thickness measuring device integrated with a temperature sensor; predicting a future thickness change trend by using a temperature-sensitive deep learning model in combination with historical working condition parameters; calculating a dynamic compensation gap sequence based on a prediction result, a target thickness and roller characteristics, and driving a roller control system to adjust the gap; synchronously collecting feedback data after rolling, adaptively updating model parameters, and forming a closed-loop control. The application provides a rolling method in a full-closed mode, high-precision rolling is realized through closed-loop control, and quality and efficiency are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent manufacturing, in particular to a rolling method, system, device and storage medium in full-closed mode. BACKGROUND

[0002] In the field of metal plate rolling processing, especially in the production of high-precision alloy plates for aerospace, semiconductor-grade foils and dissimilar metal composite plates, thickness precision is a core quality indicator. During rolling, plate thickness is affected by multiple factors such as temperature changes (affecting thermal expansion coefficient and material yield strength), mechanical properties of the press roller (such as roller gap deviation caused by thermal expansion), and material differences. Traditional processing methods face key technical challenges.

[0003] Existing technologies mainly adjust the press roller gap through artificial experience or fixed models based on physical formulas, which can meet the production needs of some conventional plates. However, in the face of temperature changes in high-speed rolling, uncertainty of new material mechanical properties, and interlayer deformation differences of dissimilar metal composite plates, traditional methods have significant limitations: first, they lack effective perception of real-time coupling effects of variables such as temperature, material, and equipment state, making it difficult to dynamically adapt to real-time changes in material thermal expansion coefficient and yield strength; second, they rely on empirical formulas or linear models for thickness prediction, which is insufficient for fitting non-linear and time-varying rolling processes, and the prediction accuracy of new materials or complex working conditions is difficult to guarantee; third, the press roller gap adjustment strategy is lagging, mostly for post-compensation or fixed time control, which cannot anticipate thickness change trends and easily leads to batch overruns; fourth, there is a lack of closed-loop calibration mechanism, and the long-term impact of external factors such as equipment thermal deformation and environmental parameters (such as humidity and air pressure) lacks adaptability, making it difficult to achieve stable control with high precision.

[0004] With the continuous improvement of plate quality requirements in high-end manufacturing, and the continuous emergence of new materials and working conditions, the shortcomings of traditional open-loop control mode in multiple implementations have become increasingly prominent. How to build a full-closed loop control system based on real-time data driving to achieve accurate prediction and anticipatory compensation of temperature-sensitive thickness changes has become a core technical problem in the field of metal plate rolling that needs to be solved. SUMMARY

[0005] The present application aims to provide a rolling method, system, device and storage medium in full-closed mode, to solve the problem of thickness precision control caused by the lack of closed-loop control capability in the plate rolling process.

[0006] One object of the present application is to provide a rolling method in full-closed mode, comprising:

[0007] Real-time thickness data and real-time temperature data of a to-be-rolled plate are acquired in real time by a thickness measurement device integrated with a temperature sensor;

[0008] A temperature-sensitive deep learning model is preset, and the real-time temperature data, the real-time thickness data and historical rolling condition parameters are taken as inputs, and a plate thickness change trend prediction matrix under different temperature evolution paths in a future preset time is output;

[0009] Based on the plate thickness change trend prediction matrix, a target thickness of the plate and current roll mechanical characteristic parameters are combined to calculate a dynamic compensation roll gap sequence, the dynamic compensation roll gap sequence including gap adjustment amounts and adjustment time sequences for a plurality of control periods in the future;

[0010] The dynamic compensation roll gap sequence is sent to a roll control system, the roll control system dynamically adjusts a roll gap, and thickness feedback data and corresponding temperature feedback data of the plate after roll adjustment and rolling are synchronously collected, and the thickness feedback data and the temperature feedback data are uploaded.

[0011] By adopting the above technical solution, real-time and synchronous collection of plate thickness and temperature data is realized by a thickness measurement device integrated with a temperature sensor, dynamic sensing capability of key parameters in the rolling process is ensured, and real-time data support is provided for accurate control. The temperature-sensitive deep learning model breaks through the limitations of traditional empirical formulas or linear models, can effectively capture the nonlinear mapping relationship between temperature evolution and plate thickness change, and significantly improves the thickness change trend prediction capability under complex conditions (such as temperature sudden change and new material rolling). The dynamic compensation roll gap sequence generated based on the prediction matrix can be adjusted in advance according to the thickness change trend in the future, compared with the traditional lagging after-compensation mechanism, the dynamic optimization and early adaptation of the roll gap are realized, and the thickness deviation caused by factors such as temperature fluctuation and material property change is effectively inhibited. By including the thickness and temperature feedback data after rolling into the closed-loop control process, a complete closed loop of “data acquisition-model prediction-gap adjustment-feedback calibration” is formed, the system can continuously optimize the control parameters according to the actual rolling effect, the self-adaptability to long-term influences such as changes in equipment state and fluctuations in environmental parameters is enhanced, and the intelligent level and thickness control precision of the metal plate rolling process are fundamentally improved, providing an innovative solution for modern rolling production with high precision and high stability.

[0012] In a possible implementation of the application, the method comprises:

[0013] A spectrum detector is integrated in the thickness measurement device integrated with a temperature sensor, plate surface spectrum data are collected in real time, and a preset material identification model is used to identify the material type and mechanical property parameters of the current rolling plate online.

[0014] According to the identified material type and mechanical property parameters, the corresponding material-specific prediction sub-model is called;

[0015] If the plate material is a new material and the historical data is insufficient, a transfer learning technology is used to adjust the temperature-sensitive deep learning model parameters based on the similar material-specific prediction sub-model combined with the mechanical property parameters;

[0016] When calculating the dynamic compensation roll gap sequence, a safety margin compensation term is added for new materials or materials with insufficient data.

[0017] By using the above technical solution, by integrating a spectral detector and a material identification model in a thickness measurement device integrated with a temperature sensor, real-time online identification of the plate material type and mechanical property parameters is realized, avoiding the lag and subjectivity caused by manual intervention, providing a data basis for accurately calling a material-specific prediction sub-model; The setting of the material-specific prediction sub-model can be customized for different material characteristics, significantly improving the thickness change trend prediction accuracy of specific materials in the rolling process compared to general models; For new materials or insufficient historical data, the application of transfer learning technology can quickly build an adaptive model based on the historical knowledge of similar materials, solving the prediction failure problem caused by data scarcity in traditional methods, significantly shortening the process debugging cycle of new materials; The addition of the safety margin compensation term enhances the fault tolerance of the system to material uncertainty, and through the compensation mechanism, the risk of thickness deviation is avoided in advance when the model prediction reliability is insufficient, thereby improving the adaptability and stability of the rolling process for new materials and complex materials, providing a universal solution for high-precision rolling of diverse materials.

[0018] In a possible implementation of the present application, the method comprises:

[0019] The humidity, air pressure and equipment vibration data of the rolling environment are obtained in real time by the environmental sensors set in the rolling site;

[0020] The influence of the humidity, air pressure and equipment vibration data on the temperature transfer and thickness change in the plate rolling process is analyzed to construct an environmental correction model;

[0021] The humidity, air pressure and equipment vibration data, the real-time temperature data, the real-time thickness data, the historical rolling process parameters, the material type and the mechanical property parameters are input into the temperature-sensitive deep learning model or the material-specific prediction sub-model;

[0022] The plate thickness change trend prediction matrix is corrected.

[0023] By adopting the technical scheme, the environmental correction model is constructed by collecting environmental parameters such as humidity, air pressure and equipment vibration in real time through the environmental sensor, and is integrated into the temperature-sensitive deep learning model, so that dynamic quantitative analysis and accurate compensation of the influence of environmental factors in the rolling process are realized: the real-time environmental data acquisition enables the system to perceive the influence of humidity on the heat dissipation efficiency of the plate surface, the effect of air pressure on the atomization effect of the cooling medium and the interference of equipment vibration on the stability of the compression roller, avoiding the prediction deviation caused by the neglect or simplified processing of environmental factors in the traditional method; the introduction of the environmental correction model quantifies the correlation between the environmental parameters and the temperature transfer and thickness change, filling the modeling gap of the coupling of multiple physical fields in a complex environment; the environmental data, material parameters and real-time working condition data are input into the intelligent model together, so that the prediction matrix can dynamically adapt to environmental changes (such as the decrease of the cooling rate of the plate in a high-humidity environment and the slight fluctuation of the compression roller gap in a high-vibration working condition), effectively correcting the thickness prediction error caused by environmental fluctuations, improving the generalization ability and prediction accuracy of the model under different environmental conditions, thereby enhancing the adaptability of the rolling system to non-ideal working conditions such as changes in workshop temperature and humidity and long-term running vibration offset of equipment, and providing environmental robustness guarantee for stable realization of high-precision thickness control.

[0024] In a possible implementation of the present application, the method comprises:

[0025] When it is monitored that the temperature change rate of the plate exceeds a preset value, an optimal stable temperature interval is calculated by the temperature-sensitive deep learning model based on the material type, the mechanical property parameters, the target thickness and the humidity, air pressure and equipment vibration data of the current plate;

[0026] The temperature of the plate is adjusted to the optimal stable temperature interval by the heating device or the cooling device integrated in the rolling line, and the adjustment power of the heating device or the cooling device is dynamically matched according to the real-time thermal radiation data of the plate;

[0027] The adjustment temperature data of the plate after temperature adjustment, the material type and the mechanical property parameters, the humidity, air pressure and equipment vibration data are synchronously input into the temperature-sensitive deep learning model to generate a low-fluctuation special prediction matrix;

[0028] The dynamic compensation compression roller gap sequence is calculated based on the low-fluctuation special prediction matrix.

[0029] By adopting the above technical solutions, after monitoring that the temperature change rate of the plate material exceeds the limit, the optimal stable temperature range is calculated based on the material characteristics, mechanical parameters and environmental data, and combined with the dynamic power adjustment of the heating / cooling device, active intervention and accurate control of the temperature violent fluctuation are realized: the temperature regulation driven by real-time thermal radiation data ensures that the plate material temperature is quickly stabilized in the range where the material deformation resistance is low and the thermal expansion coefficient is stable, avoiding the interference of yield strength mutation and uneven thermal expansion caused by temperature sudden change on thickness control; the stable temperature data and environmental and material parameters are input into the model to generate a low-fluctuation special prediction matrix, effectively filtering the noise introduced by temperature high-frequency fluctuation, so that the prediction matrix more accurately reflects the deformation law of the material in the stable thermal state; the dynamic compensation roll gap sequence calculated based on the matrix can accurately match the material flow characteristics after temperature stabilization, significantly reducing the thickness prediction deviation and compensation lag caused by unstable temperature, improving the thickness control accuracy of heat-sensitive materials such as high-temperature alloys and titanium alloys during high-speed rolling, providing a closed-loop control strategy to solve the plate deformation uneven problem under the condition of temperature sudden change, and enhancing the adaptability and stability of the rolling system to extreme temperature changes.

[0030] In a possible implementation of the present application, the method comprises:

[0031] When the rolled plate material is a dissimilar metal clad plate material, an ultrasonic flaw detector and a metallographic microscope probe are additionally arranged in the thickness measuring device of the integrated temperature sensor to obtain real-time bonding condition data of the dissimilar metal clad plate material interface, the bonding condition data including bonding strength information and interface defect information;

[0032] The spectrum of each layer of metal of the dissimilar metal clad plate material is analyzed, and the material type and mechanical property parameters of each layer of metal are accurately identified through a preset material identification model;

[0033] A comprehensive mechanical property model of the dissimilar metal clad plate material is established, and the interaction between each layer of metal is considered to more accurately describe the overall mechanical behavior of the dissimilar metal clad plate material;

[0034] According to the material type, the mechanical property parameters and the bonding condition data of each layer of metal, a special temperature-sensitive deep learning sub-model is customized, and the temperature-sensitive deep learning sub-model focuses on the heat transfer and deformation coordination relationship between different metal layers during the training process to improve the prediction accuracy of the thickness change trend of the dissimilar metal clad plate material;

[0035] In calculating the dynamic compensation roll gap sequence, the target thickness of the plate and the mechanical property parameters of the roll are considered, and in combination with the bonding condition data and the comprehensive mechanical property model of the dissimilar metal clad plate, additional gap compensation is added for the weakly bonded interface area to ensure uniform deformation and good bonding of each layer of metal during rolling.

[0036] By adopting the above technical scheme, for the interlayer coordination problem of dissimilar metal clad plate, the real-time and accurate monitoring of the interface bonding strength and defects is realized by adding ultrasonic flaw detector and metallographic microscope probe, avoiding the interlayer peeling risk caused by unknown interface state in traditional methods; the combination of spectral analysis and material identification technology ensures the accurate acquisition of the material characteristics and mechanical parameters of each layer of metal, providing data support for building a comprehensive mechanical property model considering interlayer interaction; the customized temperature-sensitive deep learning sub-model focuses on interlayer heat transfer and deformation coordination, breaking through the simplification limitation of traditional models in describing the overall mechanical behavior of the clad plate, and significantly improving the prediction accuracy of thickness change trend; the additional gap compensation for the weakly bonded interface area can dynamically balance the deformation resistance difference of each layer of metal (such as interlayer stress caused by different thermal expansion coefficients and yield strengths), promote uniform deformation of each layer of metal during rolling, effectively reduce interface cracks, peeling and other defects caused by interlayer stress concentration, and ensure the interface bonding quality and overall dimensional accuracy of the clad plate, providing a targeted solution for high-precision manufacturing of dissimilar metal clad plate.

[0037] In a possible implementation of the present application, the method comprises:

[0038] The temperature change data of each layer of metal is obtained in real time by a plurality of high-precision temperature sensors arranged at the rolling site;

[0039] The thermal expansion amount at different positions is calculated according to the thermal expansion coefficients of each layer of metal, and a thermal expansion distribution map is generated;

[0040] According to the thermal expansion amount difference of each layer of metal in the thermal expansion distribution map, the metal layer that needs to be temperature controlled and the control direction thereof are determined;

[0041] The heating device or cooling device integrated on the rolling line is used to independently control the temperature of each layer of metal;

[0042] The thermal expansion amount change of each layer of metal is monitored in real time, and the heating or cooling rate is adjusted according to the thermal expansion amount change;

[0043] Thermal expansion monitoring data of each layer of metal is obtained;

[0044] Temperature control parameters of the heating device or cooling device are obtained;

[0045] The thermal expansion monitoring data and the temperature regulation parameter are input into the temperature-sensitive deep learning sub-model, and the plate thickness change trend prediction matrix is updated in real time to more accurately reflect the influence of thermal expansion difference on the thickness and interface bonding of the dissimilar metal composite plate.

[0046] By adopting the above technical solution, for the deformation coordination problem caused by the interlayer thermal expansion difference of the dissimilar metal composite plate, the temperature change of each metal layer is captured in real time by the high-precision temperature sensor, and the thermal expansion coefficient calculation and distribution visualization are combined to realize the accurate quantitative analysis of the interlayer thermal deformation difference; based on the independent temperature regulation strategy of the thermal expansion distribution map, the temperature of each metal layer can be adjusted (such as cooling the metal layer with large thermal expansion amount and heating the layer with small expansion amount), the interlayer thermal expansion amount is dynamically balanced, and the interface shear stress caused by the expansion difference is effectively reduced; the change of the thermal expansion amount is monitored in real time and fed back to the temperature-sensitive deep learning sub-model, so that the prediction matrix can reflect the change of the mechanical behavior after the interlayer thermal state adjustment in real time, avoiding the prediction deviation caused by ignoring the interlayer heat transfer dynamic process in the traditional method; the mechanism promotes the deformation uniformity of each metal layer in the rolling process, reduces the cracks, peeling and other defects caused by thermal stress concentration in the interface bonding area, improves the overall thickness precision and interface bonding quality of the composite plate, and provides key technical support for the interlayer thermal coordination control of high-precision and high-reliability manufacturing of dissimilar metal composite plates.

[0047] In a possible implementation of the present application, the method comprises:

[0048] A distributed fiber Bragg grating sensor is embedded in the press roller body and bearing seat to collect temperature distribution data of the press roller in the circumferential and axial directions in real time, and obtain the temperature gradient of the press roller surface and the bearing seat temperature rise rate;

[0049] A press roller thermal deformation prediction model is constructed in combination with the thermal expansion coefficient of the press roller material, and based on the heat conduction equation and the expansion formula, the gap deviation caused by the radial expansion of the press roller due to temperature change is calculated;

[0050] The temperature distribution data, the gap deviation, the real-time thickness data and the real-time temperature data are input into the temperature-sensitive deep learning model to generate a thermal compensation dynamic gap sequence;

[0051] The thickness feedback data of the rolled plate is collected by the thickness measuring device, if the thickness deviation exceeds the preset range within a continuous preset period, and the deviation direction is consistent with the predicted direction of the hot deformation of the compression roller, the actual temperature of the compression roller at the current time, the actual gap between the compression roller and the plate, and the actual thickness of the plate are extracted, and the hot deformation weight parameter of the temperature-sensitive deep learning model is updated, so as to realize the adaptive optimization of the compensation strategy.

[0052] By adopting the above technical scheme, the accurate monitoring and adaptive compensation control of the hot deformation of the compression roller are realized. The distributed fiber Bragg grating sensor can collect the temperature distribution data of the compression roller in the circumferential direction and the axial direction in real time, can accurately capture the temperature gradient of the surface of the compression roller and the temperature rise rate of the bearing seat, and provides a basis for constructing an accurate hot deformation prediction model of the compression roller. The model constructed based on the heat conduction equation and the expansion formula can calculate the gap deviation caused by the radial expansion of the compression roller, so that the system can predict the influence of the hot deformation on the rolling gap in advance. The temperature distribution, gap deviation and other data are input into the temperature-sensitive deep learning model to generate a hot compensation dynamic gap sequence, so that the gap of the compression roller can be adjusted in time to compensate for the hot deformation, and the control accuracy of the rolling thickness is improved. When the thickness deviation of the rolled plate exceeds the preset range within a continuous preset period and is consistent with the predicted direction of the hot deformation of the compression roller, the related actual data are extracted to update the hot deformation weight parameter of the model, so that the adaptive optimization of the compensation strategy can be realized, the system can maintain good thickness control effect in different working conditions and long-time operation, the stability and reliability of the rolling process are enhanced, the waste rate caused by the hot deformation of the compression roller is reduced, and the product quality and production efficiency are improved.

[0053] The second object of the present application is to provide a rolling system in full-closed mode, which comprises:

[0054] A thickness and temperature data acquisition module: real-time thickness data and real-time temperature data of the to-be-rolled plate are acquired by a thickness measuring device integrated with a temperature sensor;

[0055] A plate thickness change trend prediction matrix output module: a temperature-sensitive deep learning model is preset, and the real-time temperature data, the real-time thickness data and historical rolling working condition parameters are taken as inputs to output a plate thickness change trend prediction matrix under different temperature evolution paths in a future preset time;

[0056] A dynamic compensation compression roller gap sequence calculation module: based on the plate thickness change trend prediction matrix, the target thickness of the plate and the current compression roller mechanical property parameters are combined to calculate a dynamic compensation compression roller gap sequence, and the dynamic compensation compression roller gap sequence includes gap adjustment amounts and adjustment time sequences in future control periods;

[0057] A pressure roller adjustment and feedback data uploading module is configured to send the dynamic compensation pressure roller gap sequence to a pressure roller control system, the pressure roller control system dynamically adjusts the pressure roller gap, synchronously collects thickness feedback data and corresponding temperature feedback data of the plate after the pressure roller adjustment and rolling, and uploads the thickness feedback data and the temperature feedback data.

[0058] By adopting the above technical solutions, the thickness measurement device integrated with the temperature sensor is used to realize real-time synchronous collection of the plate thickness and temperature data, ensure the dynamic perception ability of the key parameters in the rolling process, and provide real-time data support for accurate control. The introduction of the temperature-sensitive deep learning model breaks through the limitations of traditional empirical formulas or linear models, can effectively capture the nonlinear mapping relationship between temperature evolution and plate thickness change, and significantly improves the thickness change trend prediction ability under complex working conditions (such as temperature sudden change and new material rolling). The dynamic compensation pressure roller gap sequence generated based on the prediction matrix can be adjusted in advance according to the thickness change trend in the future multiple periods. Compared with the traditional lagging compensation mechanism, the dynamic optimization and early adaptation of the pressure roller gap are realized, and the thickness deviation caused by factors such as temperature fluctuation and material property change is effectively inhibited. By incorporating the thickness and temperature feedback data after rolling into the closed-loop control process, a complete closed loop of “data collection-model prediction-gap adjustment-feedback calibration” is formed, which enables the system to continuously optimize the control parameters according to the actual rolling effect, enhances the self-adaptability to long-term influences such as changes in equipment state and fluctuations in environmental parameters, and fundamentally improves the intelligent level and thickness control precision of the metal plate rolling process, providing an innovative solution for modern rolling production with high precision and high stability.

[0059] The third object of the present application is to provide a full-closed mode rolling equipment, which comprises:

[0060] a memory and a processor, wherein the memory stores a computer program capable of being loaded and executed by the processor to perform the full-closed mode rolling method.

[0061] The fourth object of the present application is to provide a storage medium.

[0062] The fourth object of the present application is achieved by the following technical solutions:

[0063] A storage medium, wherein the storage medium stores a computer program capable of being loaded and executed by the processor to perform the full-closed mode rolling method.

[0064] In summary, the present application includes at least one of the following beneficial technical effects:

[0065] 1. The thickness measurement device with integrated temperature sensors enables real-time synchronous acquisition of plate thickness and temperature data, ensuring dynamic sensing capability of key parameters during rolling, and providing real-time data support for precise control. The introduction of temperature-sensitive deep learning models breaks the limitations of traditional empirical formulas or linear models, effectively capturing the nonlinear mapping relationship between temperature evolution and plate thickness changes, significantly improving the thickness change trend prediction capability under complex conditions such as sudden temperature changes and new material rolling. The dynamic compensation roll gap sequence generated based on the prediction matrix can be adjusted in advance according to the future thickness change trend, compared with the traditional lagging post-compensation mechanism, realizing the dynamic optimization and advance adaptation of the roll gap, effectively suppressing the thickness deviation caused by factors such as temperature fluctuations and material property changes. By incorporating the post-rolling thickness and temperature feedback data into the closed-loop control process, a complete closed loop of "data acquisition-model prediction-gap adjustment-feedback calibration" is formed, enabling the system to continuously optimize control parameters based on actual rolling results, enhancing the adaptability to long-term influences such as equipment state changes and environmental parameter fluctuations, and fundamentally improving the intelligent level and thickness control precision of the metal plate rolling process, providing an innovative solution for modern rolling production with high precision and stability.

[0066] 2. For the interlayer coordination problem of dissimilar metal composite plates, by adding ultrasonic flaw detectors and metallographic microscope probes, real-time and accurate monitoring of interface bonding strength and defects is realized, avoiding the risk of interlayer peeling caused by unknown interface state in traditional methods. The combination of spectral analysis and material identification technology ensures accurate acquisition of the material properties and mechanical parameters of each layer of metal, providing data support for building a comprehensive mechanical performance model considering interlayer interaction. The customized temperature-sensitive deep learning sub-model focuses on interlayer heat transfer and deformation coordination, breaking through the simplification limitations of traditional models in describing the overall mechanical behavior of composite plates, and significantly improving the prediction accuracy of thickness change trends. The additional gap compensation amount designed for weakly bonded areas of the interface can dynamically balance the deformation resistance differences of each layer of metal (such as interlayer stress caused by differences in thermal expansion coefficient and yield strength), promoting uniform deformation of each layer of metal during rolling, effectively reducing defects such as interface cracks and peeling caused by interlayer stress concentration, and ensuring the interface bonding quality and overall dimensional accuracy of the composite plate, providing a targeted solution for high-precision manufacturing of dissimilar metal composite plates. BRIEF DESCRIPTION OF DRAWINGS

[0067] Figure 1 is a flowchart of a full-closed mode rolling method provided by an embodiment of the present application;

[0068] Figure 2 is a virtual structure schematic diagram of a full-closed mode rolling system provided by an embodiment of the present application. DETAILED DESCRIPTION

[0069] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0070] In addition, the term "and / or" in the present application is only used to describe the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B can mean that there are three cases of A alone, A and B together, and B alone. In addition, the character " / " in the present application generally represents an "or" relationship between the front and rear associated objects unless otherwise specified.

[0071] The embodiments of the present application will be described in further detail below with reference to the drawings of the specification.

[0072] The embodiments of the present application provide a rolling method in full-closed mode, referring to Figure 1 , the main process of the method is described as follows:

[0073] S1: Real-time thickness data and real-time temperature data of a corresponding position of a plate to be rolled are obtained in real time by a thickness measurement device integrated with a temperature sensor;

[0074] Among them, the thickness measurement device integrated with a temperature sensor is not a simple independent device, but realizes data synchronization through hardware integration and algorithm calibration. For example, in a certain cold rolling mill, the thickness gauge measures the thickness by laser triangulation, and the built-in infrared array sensor synchronously collects the temperature of the corresponding position. The two are strictly aligned through time stamp and space coordinate, ensuring that each thickness data point matches the real-time temperature value. This three-dimensional data matrix of "position-temperature-thickness" provides an input with physical meaning for subsequent model analysis, avoiding the data misplacement problem caused by traditional separate collection.

[0075] S2: A temperature-sensitive deep learning model is pre-set, taking the real-time temperature data, the real-time thickness data and historical rolling working condition parameters as inputs, and outputting a plate thickness change trend prediction matrix under different temperature evolution paths in a future preset time;

[0076] The core advantage of the temperature-sensitive deep learning model is to handle the nonlinear and time-varying rolling process. When the plate temperature rises rapidly from 800℃ to 850℃, the traditional linear model assumes that the thermal expansion coefficient is constant, and predicts that the thickness increases by 0.02mm; while the model is trained through historical data to find that the decrease of yield strength in the high temperature zone will cause an additional 0.01mm thickness thinning, and the final output is a more accurate prediction matrix. This capture of the complex relationship between "temperature-material performance-deformation" relies on the model's learning of a large amount of historical working conditions (such as different rolling speed and reduction rate combinations), achieving full-scenario coverage of future multi-path temperature evolution (such as uniform heating and step cooling).

[0077] S3: Based on the plate thickness change trend prediction matrix, the target thickness of the plate and the current mechanical characteristic parameters of the press roll are combined to calculate a dynamic compensation press roll gap sequence, the dynamic compensation press roll gap sequence includes gap adjustment amount and adjustment timing for future control periods;

[0078] Wherein, the generation of dynamic compensation press roll gap sequence is not single dependent on prediction result, but intelligent decision fusion of equipment physical characteristics. For example, when the model predicts that the edge of the plate will thicken by 0.03mm, the system needs to consider the mechanical stiffness of the press roll at the same time, if the central stiffness of the press roll is higher than that of the two ends, the actual reduction effect of the same gap adjustment amount in the edge area is more significant, therefore the gap sequence will automatically adjust the compensation amount of the edge area (such as increasing 0.005mm), to ensure that the theoretical calculation is consistent with the actual response of the equipment. This "model prediction + equipment characteristics" coupled calculation avoids the problem of disconnection between ideal model and actual equipment in traditional methods.

[0079] S4: Send the dynamic compensation press roll gap sequence to the press roll control system, the press roll control system dynamically adjusts the press roll gap, synchronously collects the thickness feedback data and the corresponding temperature feedback data of the plate after the press roll is adjusted and the rolling is completed, and uploads the thickness feedback data and the temperature feedback data.

[0080] Wherein, when the dynamic compensation roll gap sequence is generated, it will be sent quickly and stably to the roll control system through industrial Ethernet. The core of the roll control system is the electro-hydraulic servo mechanism, which has high response speed and precise position control capability. Taking the common four-high rolling mill as an example, when receiving the gap adjustment instruction, the electro-hydraulic servo valve will quickly adjust the flow and pressure of hydraulic oil to drive the press screw or hydraulic cylinder to move the roll position accurately, ensuring that the gap is adjusted in place. After the roll completes the gap adjustment and rolls the plate, the thickness measuring device integrated with the temperature sensor will play a role again. It will immediately measure the rolled plate and synchronously collect the thickness feedback data and the corresponding temperature feedback data. The collection accuracy of the thickness feedback data is crucial, for example, using laser interference thickness measurement technology can control the measurement accuracy within ±0.005mm. At the same time, the temperature sensor will record the temperature of the corresponding position of the plate, such as the middle temperature of the plate being 55℃. These feedback data will be uploaded to the database of the control system in real time, providing a basis for subsequent analysis and calibration.

[0081] The system will analyze the uploaded feedback data in real time. If the thickness deviation occurs in multiple consecutive periods (such as 5 consecutive periods) and the deviation direction is consistent with the prediction direction of the temperature-sensitive deep learning model before, it indicates that the prediction of the model may have some errors and needs to be calibrated. At this time, the system will extract the "roll temperature-actual gap-plate thickness" data pair at the current time. For example, the middle temperature of the roll is 60℃, the actual gap is 2.55mm, and the measured thickness of the middle of the plate is 2.52mm.

[0082] Then, the system will update the weight parameters of the temperature-sensitive deep learning model using online learning algorithms (such as stochastic gradient descent algorithm). Online learning algorithms can adjust the model in real time according to newly collected data, so that the model can quickly adapt to various changes in the rolling process. By continuously updating the model parameters, the prediction accuracy of the subsequent prediction will be optimized, thereby realizing the complete closed-loop control of prediction-adjustment-feedback-calibration. This closed-loop control mechanism can effectively improve the stability of the rolling process and the product quality, reduce the thickness deviation caused by various factors (such as material property fluctuations, environmental changes, equipment wear, etc.), and ensure that the thickness of the plate is always controlled within the target range.

[0083] In another embodiment, for the full-closed mode rolling method, multi-dimensional data fusion and intelligent strategy upgrading can be used to further improve the technical innovation, which is suitable for harsh scenarios such as high-precision electronic foil and aerospace ultra-thin plate:

[0084] First, high-density sensor arrays are introduced in the data acquisition link. In addition to real-time thickness and temperature detection, the surface stress distribution and micro-area displacement data are also collected synchronously, forming a multi-modal data set of "thickness-temperature-stress-displacement". For example, in the rolling of ultra-thin copper foil, detection points are arranged at micron-level intervals along the width direction of the plate to accurately capture the material softening area caused by local friction overheating, providing more rich physical feature input for subsequent prediction and solving the problem of micro-deformation difference that cannot be identified by traditional single-point detection.

[0085] Second, the temperature-sensitive model is upgraded to an adaptive learning framework, which integrates multiple data sources such as equipment vibration signals and hydraulic system pressure fluctuations to build a "prediction-decision-verification" closed-loop optimization mechanism. The model not only outputs the thickness change trend, but also dynamically calculates the stress safety threshold of each region. When the edge area stress approaches the damage threshold, the local roll gap is automatically adjusted to ensure thickness accuracy while avoiding plate tearing, significantly improving control robustness under complex conditions.

[0086] In the dynamic compensation strategy, digital twin pre-validation technology is introduced to simulate the rolling effect of different gap adjustment schemes based on the physical model of the roll and the plate. For example, when the system generates a gap adjustment sequence, it first verifies whether the scheme will cause edge thickness to exceed the tolerance or stress concentration through a virtual model. If there is a risk, the adjustment amplitude and timing are automatically optimized. This converts the traditional physical trial-and-error of trial rolling into pre-validation in the digital space, ensuring the safety and accuracy of the control scheme, especially for zero-risk commissioning of new materials.

[0087] Finally, edge computing technology is used to realize real-time processing of feedback data locally, with thickness deviation analysis and equipment state monitoring tasks sinking to the mill site node to shorten data transmission and response delay. For example, when an abnormal high-frequency vibration signal is detected, the edge node can quickly identify it as an early feature of roll bearing wear, triggering a maintenance warning and adaptively adjusting control parameters to improve the system's response speed to equipment aging, environmental fluctuations, and other disturbances to sub-millisecond level, further enhancing the accuracy and stability of long-term operation.

[0088] Through the expansion of detection dimensions, intelligent upgrading of models, virtual-real fusion verification, and edge computing enhancement, the whole process of intelligent upgrading from data acquisition to control execution is formed, effectively solving the problems of insufficient micro-deformation monitoring, poor adaptability to extreme conditions, and equipment response lag in high-end plate rolling.

[0089] Specifically, in some possible embodiments, the method comprises:

[0090] In the thickness measurement device integrated with the temperature sensor, a spectral detector is integrated to collect real-time spectral data of the plate surface, and a pre-set material identification model is used to identify the material type and mechanical property parameters of the current rolling plate online.

[0091] According to the identified material type and mechanical property parameters, the corresponding material-specific prediction sub-model is called.

[0092] If the plate is a new material and the historical data is insufficient, the transfer learning technology is used to adjust the temperature-sensitive deep learning model parameters based on the similar material-specific prediction sub-model combined with the mechanical property parameters.

[0093] When calculating the dynamic compensation roll gap sequence, a safety margin compensation term is added for new materials or materials with insufficient data.

[0094] Among them, in the thickness measurement device integrated with the temperature sensor, when the plate to be rolled enters the detection area, the spectral detector will collect the spectral data of the plate surface in real time. For example, in the rolling workshop, the spectral detector collects spectral information of the plate surface at a frequency of 10 times per second. The collected spectral data is transmitted to the control system for analysis by a pre-set material identification model. This material identification model is trained based on the spectral characteristics of a large number of known materials, which can identify the material type of the current rolling plate online, such as carbon steel, aluminum alloy or stainless steel, etc., and can also estimate the mechanical property parameters of the material, such as elastic modulus and yield strength. For example, for a plate to be rolled, the material identification model identifies its material as 304 stainless steel and its elastic modulus as about 193 GPa and its yield strength as about 205 MPa by analyzing the spectral data.

[0095] According to the material type and mechanical property parameters identified by the material identification model, the control system will call the corresponding material-specific prediction sub-model from the pre-established model library. The model library stores specific prediction sub-models for different common materials, which are trained based on a large amount of historical rolling data of the material and can more accurately predict the thickness change trend of the material during rolling. For example, if the identified plate material is 6061 aluminum alloy, the control system will call the 6061 aluminum alloy-specific prediction sub-model. This sub-model will take real-time temperature data, real-time thickness data and historical rolling working condition parameters as input, and output a plate thickness change trend prediction matrix that is more suitable for the characteristics of the material.

[0096] In actual production, new material plates or materials with insufficient historical data may be encountered. When such a situation is encountered, the parameters of the temperature-sensitive deep learning model are adjusted using transfer learning technology. First, the control system searches for a similar material-specific prediction sub-model in the model library. For example, for a new titanium alloy plate, since it has insufficient historical data, the control system finds that it is similar to the common TC4 titanium alloy in composition and performance, so it selects the TC4 titanium alloy-specific prediction sub-model as the basis. Then, the parameters of the temperature-sensitive deep learning model are adjusted in combination with the mechanical property parameters of the new material identified through spectral data. Through transfer learning, the model can quickly adapt to the characteristics of the new material, and although the new material has limited historical data, it can still accurately predict the thickness change trend during rolling to some extent.

[0097] When calculating the dynamic compensation roll gap sequence, a safety margin compensation term needs to be added for new materials or materials with insufficient data. This is to reduce the risk of rolling due to insufficient understanding of the characteristics of new materials. For example, for the new titanium alloy plate mentioned above, when calculating the dynamic compensation roll gap sequence, a safety margin compensation term is added to the originally calculated gap adjustment amount, such as increasing the roll gap by 0.05mm. This can avoid the problem of over-pressing or excessive thickness deviation of the plate during rolling due to inaccurate prediction of the deformation characteristics of the new material. In the subsequent rolling process, as the amount of data collected for the new material gradually increases, the safety margin compensation term can be dynamically adjusted according to the actual situation to achieve more accurate rolling control.

[0098] Specifically, in some possible embodiments, the method comprises:

[0099] Real-time acquisition of humidity, air pressure, and equipment vibration data of the rolling environment through environmental sensors set in the rolling field;

[0100] Analysis of the influence of the humidity, air pressure, and equipment vibration data on temperature transfer and thickness change during plate rolling to construct an environmental correction model;

[0101] Inputting the humidity, air pressure, and equipment vibration data, the real-time temperature data, the real-time thickness data, and the historical rolling working condition parameters, the material type, and the mechanical property parameters into the temperature-sensitive deep learning model or the material-specific prediction sub-model;

[0102] Correcting the plate thickness change trend prediction matrix.

[0103] In actual metal plate rolling production, the influence of environmental factors on plate thickness control is often complex and diverse. Changes in humidity and air pressure can change the heat dissipation efficiency of the material, and equipment vibration can cause dynamic fluctuations in the gap between the rolls. To accurately capture these influences, multiple environmental sensors are deployed at key locations in the rolling field: high-precision temperature and humidity transmitters are installed near the entrance of the rolling mill to collect air humidity and air pressure data in real time, such as monitoring humidity of 65% and air pressure of 101.3kPa at a frequency of 100 times per second; three-axis vibration acceleration sensors are fixed on the roll bearing seat and the rack to capture the amplitude and frequency of equipment vibration in real time, for example, detecting a periodic vibration signal of 150Hz, 0.3g. These sensors are connected in real time with the control system through an industrial network, ensuring that the environmental data is completely synchronized in time with the thickness and temperature data of the plate, providing reliable raw data for subsequent analysis.

[0104] After collecting environmental data, the system uses historical rolling data and machine learning algorithms to analyze the correlation between humidity, air pressure, equipment vibration, and plate thickness changes. For example, through a large number of samples, it is found that for every 10% increase in humidity, the heat dissipation rate of the plate surface to the air decreases by 5%, resulting in a 0.5°C / s decrease in plate temperature under the same cooling conditions; and for every 0.1g increase in the amplitude of equipment vibration in the 100-200Hz frequency band, the actual fluctuation range of the roll gap expands by 0.002mm, directly affecting the stability of thickness measurement. Based on these rules, the system builds an environmental correction model that converts humidity, air pressure, vibration, and other parameters into correction factors for key physical quantities such as temperature transfer efficiency and equipment stiffness, such as generating a humidity correction factor K_h=0.95 indicating a 5% decrease in heat dissipation efficiency under the current humidity, and a vibration correction factor K_v=1.02 indicating a 2% equivalent decrease in roll stiffness due to vibration.

[0105] These environmental correction factors do not act independently, but are input into a temperature-sensitive deep learning model along with real-time temperature, thickness, and material parameters. The input layer of the model thus expands the environmental parameter dimension, forming a comprehensive input vector containing more than ten features such as material type, mechanical properties, temperature, thickness, humidity, air pressure, and vibration. When calculating the thickness change trend of the plate, the model automatically calls the output results of the environmental correction model to adjust the temperature evolution path and material deformation law. For example, when rolling aluminum alloy plates in a high-humidity environment, the model will identify the slow heat dissipation caused by humidity, thereby prolonging the duration of the high-temperature state of the plate, and accordingly adjusting the prediction of the decrease in material yield strength, making the calculation of the influence of temperature on thickness change more accurate.

[0106] The integration of environmental data directly affects the generation of dynamic compensation roll gap sequence. When the environmental correction model detects abnormal equipment vibration or significant humidity changes, the system will increase targeted compensation measures in the original gap adjustment strategy. For example, when the vibration acceleration exceeds 0.5g, in order to offset the risk of gap fluctuation caused by vibration, an additional safety margin of 0.003mm will be added to the calculated gap adjustment amount; when the humidity is 20% higher than the baseline value, considering that the decreased heat dissipation efficiency may cause the plate temperature to be too high and the material to soften, the gap adjustment amount in the middle region of the roll will be reduced in advance to avoid excessive thinning due to the enhanced deformation ability of the material at high temperature. This mechanism of converting environmental influences into control strategies in real time enables the system to actively adapt to subtle changes in the workshop environment rather than relying on fixed empirical parameters.

[0107] Taking the rolling process of a certain stainless steel plate in a high humidity environment as an example, when the sensor detects that the humidity reaches 75% (15% higher than the baseline value), the air pressure decreases slightly, and the equipment vibration is at a moderate level, the environmental correction model first calculates that the heat dissipation efficiency decreases by 12%, which means that the cooling speed of the plate during rolling will be slower than the ideal state, and the temperature may be 3℃ higher than expected. The temperature-sensitive model adjusts the thickness change prediction accordingly, and additionally considers the 0.008mm thinning amount caused by the decrease in material yield strength due to high temperature in the original calculation. In the generated dynamic compensation gap sequence, the gap adjustment amount in the middle region is reduced from the initial 0.05mm to 0.042mm to offset the excessive thinning caused by high temperature, while the global safety margin is increased by 0.003mm to address the vibration risk. The final measured results show that the plate thickness deviation is significantly narrowed from ±0.03mm of the traditional method to ±0.015mm, fully demonstrating the actual effect of environmental factor correction on improving rolling precision. Through this way of deeply integrating environmental data into the control process, the rolling system is no longer limited by fixed environmental assumptions, but can dynamically adjust the strategy according to the real-time changes in the workshop conditions, fundamentally enhancing the adaptability to complex production environments.

[0108] Specifically, in some possible embodiments, the method comprises:

[0109] When the temperature change rate of the plate exceeds the pre-value, based on the material type of the current plate, the mechanical property parameters, the target thickness, and the humidity, air pressure, and equipment vibration data, the optimal stable temperature interval is calculated by the temperature-sensitive deep learning model;

[0110] The temperature of the plate is adjusted to the optimal stable temperature interval by the heating device or cooling device integrated in the rolling line, and the adjustment power of the heating device or cooling device is dynamically matched according to the real-time thermal radiation data of the plate;

[0111] The adjusted temperature data after the plate temperature is adjusted, the material type, and the mechanical property parameters, humidity, air pressure, and equipment vibration data are input into the temperature-sensitive deep learning model to generate a low fluctuation exclusive prediction matrix;

[0112] The dynamic compensation roll gap sequence is calculated based on the low fluctuation exclusive prediction matrix.

[0113] Wherein, in the actual metal plate rolling process, the sharp change of temperature will significantly affect the mechanical properties of the material, and then lead to the increase of thickness control difficulty. When the system monitors the temperature change rate of the plate through the integrated temperature sensor and exceeds the preset threshold (for example, the temperature rise or drop exceeds 10℃ per second), a targeted temperature stabilization control mechanism will be triggered. Taking the hot rolling process of a certain aerospace titanium alloy plate (target thickness 1.5mm) as an example, when it is detected that the plate surface temperature rises from 800℃ to 850℃ within 50ms (change rate 100℃ / s), which is far beyond the preset safety threshold of 20℃ / s, the system immediately starts the following process:

[0114] Firstly, the temperature-sensitive deep learning model will combine the current material type (titanium alloy TC4), mechanical property parameters (thermal expansion coefficient 8.6×10⁻ 6 / ℃, yield strength 850MPa), target thickness, and real-time environmental data (humidity 60%, air pressure 101kPa, equipment vibration acceleration 0.2g) to calculate the optimal stable temperature interval suitable for the material. The model analyzes the historical data and finds that TC4 titanium alloy has a thermal expansion coefficient fluctuation of less than 5% and a yield strength change rate of less than 3% in the range of 820-840℃, which is the most stable temperature range for material deformation resistance in the rolling process. Therefore, the system sets 820-840℃ as the optimal stable temperature interval for the current working condition to avoid the problem of material softening and uneven thermal expansion caused by continuous temperature rise.

[0115] Next, the heating / cooling device integrated with the rolling line dynamically adjusts the power according to the real-time thermal radiation data of the plate. The infrared thermal imager installed above the rolling mill scans the plate surface temperature distribution in real time, and finds that the current plate middle temperature has reached 860°C, the edge temperature is 830°C, and there is a obvious temperature gradient. The system calculates the required adjustment power for each region accordingly: for the high-temperature middle region, the high-pressure air cooling device is started, and the cooling air volume is increased from 500m³ / h to 800m³ / h, so that the middle temperature decreases at a rate of 20°C / s; for the low-temperature edge region, the infrared heating lamp is turned on, and the power is increased from 20kW to 30kW, to ensure that the edge temperature is maintained at 830°C. Through this dynamic power matching, the overall temperature of the plate stabilizes in the interval of 825-835°C within 200ms, meeting the optimal stable temperature requirement.

[0116] When the plate temperature adjustment stabilizes, the new adjustment temperature data (such as 830°C in the middle and 825°C in the edge) are input into the temperature-sensitive deep learning model synchronously with material parameters and environmental data. At this time, the "low fluctuation special prediction matrix" generated by the model is no longer disturbed by temperature sudden changes, and can more accurately reflect the deformation law of the material in the stable thermal state. For example, the model predicts that the thermal expansion amount difference of each region of the plate in the stable temperature interval will decrease from 0.05mm during the sudden change to within 0.01mm, and the thickness change risk caused by the fluctuation of yield strength will be reduced by 60%. This prediction matrix based on stable temperature provides a more reliable basis for subsequent gap adjustment.

[0117] Finally, when calculating the dynamic compensation roll gap sequence based on the low fluctuation special prediction matrix, the system does not need to consider the uncertainty caused by the dramatic change of temperature. For example, for the slight thickening trend (predicted thickening 0.015mm) of the edge region of the titanium alloy plate at the stable temperature, the system only needs to reduce the edge gap of the compression roller by 0.01mm to accurately compensate for the thickness deviation, avoiding the problem of over-adjustment or adjustment lag caused by temperature fluctuations in traditional methods. The measured data shows that after adopting this mechanism, the thickness deviation of the titanium alloy plate under the condition of temperature sudden change is stabilized within ±0.01mm from ±0.04mm, effectively solving the problem of uneven deformation of heat-sensitive materials during high-speed rolling caused by unstable temperature, and improving the yield rate and reliability of high-end plates.

[0118] In particular, in some possible embodiments, the method comprises:

[0119] When the rolled plate is a dissimilar metal composite plate, an ultrasonic flaw detector and a metallographic microscope probe are added to the thickness measuring device of the integrated temperature sensor to obtain real-time bonding condition data of the dissimilar metal composite plate interface, the bonding condition data including bonding strength information and interface defect information;

[0120] respectively, the material type and mechanical property parameters of each layer of metal of the dissimilar metal composite plate are accurately identified through a pre-set material identification model;

[0121] Meanwhile, a comprehensive mechanical property model of the dissimilar metal composite plate is established, and the interaction between each layer of metal is considered to more accurately describe the overall mechanical behavior of the dissimilar metal composite plate;

[0122] According to the material type, mechanical property parameters and bonding condition data of each layer of metal, a temperature-sensitive deep learning sub-model is customized, which focuses on the heat transfer and deformation coordination relationship between different metal layers during training to improve the prediction accuracy of the thickness change trend of the dissimilar metal composite plate;

[0123] When calculating the dynamic compensation roll gap sequence, the target thickness of the plate and the roll mechanical property parameters are considered, and the bonding condition data and the comprehensive mechanical property model of the dissimilar metal composite plate are combined to increase the additional gap compensation amount for the weakly bonded interface area to ensure uniform deformation and good bonding of each layer of metal of the dissimilar metal composite plate during rolling.

[0124] Among them, when actually rolling dissimilar metal composite plates (such as aluminum / steel composite plates, titanium / copper composite plates), the hardware and algorithm need to be upgraded based on the existing detection device to meet the special needs of interlayer bonding and deformation coordination. An ultrasonic flaw detector and a metallographic microscope probe are added to the thickness measurement device integrated with a temperature sensor. The ultrasonic flaw detector scans the plate interface at a frequency of 50 times per second, analyzes the bonding strength and identifies interface defects (such as cracks, inclusions, with a minimum detection size of 50 μm) through reflected echo signals; the metallographic microscope probe real-time acquires interface microstructure images to detect interlayer metallurgical bonding state (such as grain interlocking degree). When the bonding strength of a certain area is detected to be only 80 MPa (lower than the critical value) or there is a micro-crack, the system immediately marks it as a "weakly bonded area" to provide accurate positioning for subsequent control.

[0125] Secondly, the composition of the upper and lower layers of metal of the composite plate is analyzed by a spectral detector. The upper aluminum layer is made of 3003 aluminum alloy (thermal expansion coefficient 23x10⁻ 6 / ℃, yield strength 110 MPa), and the lower steel layer is Q235 (thermal expansion coefficient 12x10⁻ 6 / ℃, yield strength 235 MPa). The material identification model independently outputs mechanical parameters according to the spectral data of each layer, and establishes a comprehensive mechanical property model. The interlayer interaction is simulated by the finite element method: when the aluminum layer temperature is 20℃ higher than the steel layer, the thermal expansion difference will generate a shear stress of 5MPa at the interface, which may lead to interlayer sliding if the bonding strength is insufficient. The model converts this interlayer stress-strain relationship into a global deformation correction coefficient, for example, the actual reduction rate of the aluminum layer needs to consider the restraint effect of the steel layer, and the correction coefficient is 0.85.

[0126] Based on the above data, the system customizes a temperature-sensitive deep learning sub-model, which adds new interface bonding strength (such as 100MPa), defect location (such as 100mm from the edge), and material parameters of each layer (aluminum layer elastic modulus 70GPa, steel layer 210GPa) to the input parameters in addition to traditional temperature, thickness, and working conditions. During model training, the focus is on learning the interlayer heat transfer delay (the time constant for the measured aluminum layer heat to conduct to the steel layer is 20ms) and deformation coordination rules (such as for every 0.01mm reduction in the aluminum layer, the steel layer is reduced by 0.006mm due to restraint). The output thickness change trend prediction matrix is no longer a global prediction, but is divided by layer (such as the aluminum layer is predicted to reduce by 0.02mm, and the steel layer is predicted to reduce by 0.015mm), and the interface stress concentration area is labeled (such as the edge 20mm range has a high risk of stress supercritical value).

[0127] In calculating the dynamic compensation roll gap sequence, the system first calculates the basic gap adjustment amount (such as reducing the overall gap by 0.1mm) based on the target thickness (total thickness 2.0mm, aluminum layer 0.45mm, steel layer 1.55mm) and the mechanical characteristics of the roll (roll stiffness in the middle > both ends). For the weakly bonded area at the interface (such as the marked edge 100mm), the system automatically increases the additional compensation amount: if the bonding strength is 80MPa (80% of the critical value), then the gap adjustment amount in this area is increased by an additional 0.005mm to reduce the interlayer shear stress (from 5MPa to 3MPa). This zoned compensation strategy balances the deformation resistance differences of each layer of metal (uneven reduction rate caused by aluminum-soft-steel-hard) through basic adjustment and weak area enhancement, avoiding tearing or peeling of the bonding surface caused by uniform adjustment of the gap.

[0128] In actual rolling process, when it is detected that the interface bonding strength of a batch of composite plates is generally low (average 90MPa), the sub-model will automatically increase the additional compensation amount in the edge area from 0.005mm to 0.01mm, and extend the low-speed stable rolling time of the roll (from 10s to 15s), to ensure the uniformity of interlayer deformation, significantly improving the manufacturing quality of dissimilar metal composite plates.

[0129] Specifically, in some possible embodiments, the method comprises:

[0130] By means of a plurality of high-precision temperature sensors arranged at the rolling site, temperature change data of each layer of metal is acquired in real time;

[0131] According to the thermal expansion coefficients of each layer of metal, the thermal expansion amounts at different positions are calculated, and a thermal expansion distribution map is generated;

[0132] According to the thermal expansion amount differences of each layer of metal in the thermal expansion distribution map, the metal layer that needs to be temperature-regulated and the regulation direction thereof are determined;

[0133] By means of the heating device or the cooling device integrated in the rolling line, each layer of metal is independently temperature-regulated;

[0134] The thermal expansion amount changes of each layer of metal are monitored in real time, and the heating or cooling rate is adjusted according to the thermal expansion amount changes;

[0135] Thermal expansion monitoring data of each layer of metal are acquired;

[0136] Temperature regulation parameters of the heating device or the cooling device are acquired;

[0137] The thermal expansion monitoring data and the temperature regulation parameters are input into the temperature-sensitive deep learning sub-model, and the plate thickness change trend prediction matrix is updated in real time, so as to more accurately reflect the influence of thermal expansion differences on the thickness and interface bonding of the dissimilar metal composite plate.

[0138] Among them, at the rolling site, a plurality of high-precision temperature sensors are installed at different key positions of the rolling mill, such as the inlet, the rolling area and the outlet. For aluminum-steel composite plates, sensors are arranged at corresponding positions of the aluminum plate layer and the steel plate layer. These sensors have high precision and fast response characteristics, and can collect temperature data of each layer of metal in real time at a frequency of 10 times per second. For example, at the beginning of rolling, the temperature of the aluminum plate layer is 200℃, and the temperature of the steel plate layer is 150℃.

[0139] The thermal expansion coefficients of the aluminum plate and the steel plate are known. According to the collected temperature change data of each layer of metal, the thermal expansion amount at different positions is calculated by means of the thermal expansion coefficient formula ΔL=L0×α×ΔT (where ΔL is the thermal expansion amount, L0 is the initial length, α is the thermal expansion coefficient, and ΔT is the temperature change amount). Assuming that the initial length of the aluminum plate and the steel plate is 1m, when the temperature of the aluminum plate increases by 50℃ and the temperature of the steel plate increases by 30℃, the thermal expansion amount of the aluminum plate is calculated to be 0.00115m, and the thermal expansion amount of the steel plate is calculated to be 0.00036m. By calculating the thermal expansion amounts at different positions of the entire plate, a thermal expansion distribution map is generated, which directly displays the difference in thermal expansion amount of each layer of metal.

[0140] When the thermal expansion distribution is analyzed and it is found that the thermal expansion difference between the aluminum plate layer and the steel plate layer is large, temperature regulation is required. If the thermal expansion of the aluminum plate is too large, it may cause the composite plate to warp, the interface to be poorly combined, etc. In this case, it is determined that the temperature of the aluminum plate layer needs to be regulated, and the regulation direction is to reduce the temperature. Conversely, if the thermal expansion of the steel plate is too small, the regulation direction is to increase the temperature.

[0141] The rolling line is integrated with heating devices and cooling devices. For the aluminum plate layer that needs to be reduced in temperature, the cooling device is started, such as using air cooling or water cooling. By adjusting the flow and temperature of the cooling medium, the cooling rate of the aluminum plate layer is accurately controlled. For example, the air cooling device is turned on, and the flow of the cooling air is set to 10 m³ / min, and the temperature is set to 20℃. For the steel plate layer that needs to be increased in temperature, the heating device is started, such as using induction heating or resistance heating. By adjusting the heating power, the heating rate of the steel plate layer is controlled.

[0142] During the temperature regulation process, the change of the thermal expansion of each metal layer is continuously monitored in real time. The length change of each layer of metal can be measured by a displacement sensor or other equipment at a frequency of 5 times per second, so as to calculate the real-time change of the thermal expansion. According to the change of the thermal expansion, the heating or cooling rate is dynamically adjusted. If it is found that the thermal expansion of the aluminum plate layer decreases too slowly, it means that the cooling rate is not enough, and the cooling air flow can be increased to 15 m³ / min. If the thermal expansion of the steel plate layer increases too fast, the heating power can be reduced to 40kW.

[0143] During the whole process, the thermal expansion monitoring data of each metal layer is continuously obtained, including the real-time value of the thermal expansion, the change rate, etc. At the same time, the temperature regulation parameters of the heating device or the cooling device are recorded, such as the heating power, the cooling medium flow and the temperature, etc. These thermal expansion monitoring data and temperature regulation parameters are input into the temperature-sensitive deep learning sub-model. The sub-model will update the plate thickness change trend prediction matrix in real time according to the newly input data. For example, originally it is predicted that the overall thickness of the composite plate will increase by 0.1mm, but after considering the real-time change of the thermal expansion difference, the predicted thickness is adjusted to increase by 0.08mm, so as to more accurately reflect the influence of the thermal expansion difference on the thickness and interface combination of the dissimilar metal composite plate. It can effectively control and manage the thermal expansion difference of each layer of metal of the dissimilar metal composite plate, and improve the stability of the rolling process and the quality of the composite plate.

[0144] Specifically, in some possible embodiments, the method comprises:

[0145] A distributed fiber Bragg grating sensor is embedded in the roll body and bearing seat of the pressure roller to collect temperature distribution data of the pressure roller in the circumferential and axial directions in real time, and to obtain the temperature gradient of the pressure roller surface and the bearing seat temperature rise rate.

[0146] In combination with the thermal expansion coefficient of the compression roller material, a compression roller thermal deformation prediction model is constructed, and based on the heat conduction equation and the expansion formula, the gap deviation caused by the radial expansion of the compression roller due to temperature change is calculated.

[0147] The temperature distribution data, the gap deviation, the real-time thickness data, and the real-time temperature data are collectively input into the temperature-sensitive deep learning model to generate a thermal compensation dynamic gap sequence.

[0148] The thickness feedback data of the rolled plate is collected by the thickness measuring device. If the thickness deviation exceeds the preset range within a continuous preset period, and the deviation direction is consistent with the predicted direction of the compression roller thermal deformation, the actual temperature of the compression roller, the actual gap between the compression roller and the plate, and the actual thickness of the plate are extracted at the current time, and the compression roller thermal deformation weight parameter of the temperature-sensitive deep learning model is updated to realize adaptive optimization of the compensation strategy.

[0149] In actual rolling mill equipment operation, the thermal deformation of the compression roller will cause the gap deviation, which will further affect the thickness accuracy of the plate rolling. Distributed fiber Bragg grating sensors are installed inside the compression roller body and bearing seat. In the compression roller body, sensors are installed at certain angles along the circumferential direction and at certain distances along the axial direction; sensors are also arranged near the outer ring of the bearing in the bearing seat. These sensors can collect data with high precision and speed, such as 100 times per second. When the compression roller rotates at high speed, the sensors will maintain real-time communication through a specific device and control system to ensure accurate acquisition of the temperature distribution of the compression roller in the circumferential and axial directions. For example, it can monitor that the temperature in the middle of the compression roller body is higher than that at both ends, and it can also know the heating speed of the bearing seat due to bearing friction.

[0150] Using the collected temperature data, in combination with the thermal expansion characteristics of the compression roller material, a model that can predict the thermal deformation of the compression roller is constructed. This model is based on the principle of heat conduction, and by simulating the temperature change inside the compression roller, it calculates the expansion degree of the compression roller in the radial direction due to temperature change. According to the expansion degree of the compression roller, the deviation value of the gap between the compression roller and the plate can be further obtained. For example, it can calculate how much the gap in a certain area of the compression roller is reduced due to thermal expansion, and thus obtain the gap deviation situation at different positions of the compression roller in the axial direction.

[0151] The temperature distribution data of the press roller, the calculated gap deviation data, and the real-time thickness and temperature data of the plate are input into the temperature-sensitive deep learning model. This model analyzes and processes these data to find the correlation between the press roller thermal deformation and the plate thickness deviation. For example, it determines that the thermal expansion of a certain area of the press roller is the main cause of the thickness deviation of the corresponding area of the plate. Then, the model generates a dynamic gap sequence for thermal compensation, which contains the adjustment strategy for the gap of different positions of the press roller in the next period of time. Through the electro-hydraulic servo system, the position of the press roller can be quickly adjusted according to this sequence to dynamically correct the gap.

[0152] After the plate rolling is completed, the thickness feedback data of the plate is collected through the thickness measuring device. If the plate thickness deviation measured in several consecutive cycles exceeds the pre-set range and the deviation direction is consistent with the previous press roller thermal deformation prediction direction, it means that the model may have errors. At this time, the system extracts the actual temperature of the press roller at the current time, the actual gap between the press roller and the plate, and the actual thickness of the plate, etc. Using these data, the weight parameters of the temperature-sensitive deep learning model related to the thermal deformation of the press roller are updated through a specific algorithm. After multiple iterations, the model can more accurately predict the impact of press roller thermal deformation on the gap, and the plate thickness deviation can be controlled within a smaller range.

[0153] In the process of high-speed rolling of stainless steel plates, the press roller generates a large amount of heat due to friction with the plate, causing the temperature of the bearing seat to rise rapidly and the temperature of the middle part of the roller body to rise significantly. If no thermal deformation compensation is performed, the thickness deviation of the plate will become larger and larger. After adopting the above scheme, the sensor can capture the thermal expansion of the press roller in time, the model can generate a corresponding thermal compensation gap sequence to adjust the press roller gap, and the model parameters can be continuously optimized through feedback calibration. Finally, the thickness deviation of the plate can be controlled within a very small range, ensuring the stability of the quality of the plate in the high-speed rolling process.

[0154] This method combines equipment thermal state monitoring, model calculation, and data-driven prediction to form a complete closed-loop control system. Through accurate temperature data collection, thermal deformation model prediction, and adaptive adjustment of the deep learning model, the system can respond in real time to the impact of press roller thermal deformation, effectively improving the accuracy and stability of plate rolling, especially suitable for high-speed and high-temperature rolling production scenarios.

[0155] Another embodiment of the present application provides a full-closed mode rolling system, wherein Figure 2 A full-closed mode rolling system comprises:

[0156] The thickness and temperature data acquisition module 100: the real-time thickness data and the real-time temperature data of the corresponding position of the plate to be rolled are acquired in real time through the thickness measurement device integrated with the temperature sensor;

[0157] The plate thickness change trend prediction matrix output module 200: a temperature-sensitive deep learning model is pre-set, and the real-time temperature data, the real-time thickness data and historical rolling condition parameters are taken as inputs to output a plate thickness change trend prediction matrix under different temperature evolution paths in a future preset time;

[0158] The dynamic compensation roll gap sequence calculation module 300: based on the plate thickness change trend prediction matrix, the target thickness of the plate and the current roll mechanical property parameters are combined to calculate a dynamic compensation roll gap sequence, the dynamic compensation roll gap sequence includes gap adjustment amounts and adjustment time sequences for a plurality of control periods in the future;

[0159] The roll adjustment and feedback data uploading module 400: the dynamic compensation roll gap sequence is sent to the roll control system, the roll control system dynamically adjusts the roll gap, synchronously collects the thickness feedback data and the corresponding temperature feedback data of the plate after the roll adjustment and the completion of rolling, and uploads the thickness feedback data and the temperature feedback data.

[0160] The rolling system in the full-closed mode provided in the embodiment can realize the steps of the foregoing embodiment and achieve the same technical effects as the foregoing embodiment due to the functions of the modules themselves and the logical connections between the modules. The principle analysis can be referred to the related description of the steps of the rolling method in the full-closed mode, which will not be repeated here.

[0161] The rolling device in the full-closed mode provided in the embodiment includes a memory and a processor, and the memory stores a computer program capable of being loaded and executed by the processor to perform the rolling method in the full-closed mode.

[0162] The rolling device in the full-closed mode provided in the embodiment includes a memory and a processor, and the memory stores a computer program capable of being loaded and executed by the processor to perform the rolling method in the full-closed mode.

[0163] The storage medium provided in the embodiment will realize the steps of the foregoing embodiment and achieve the same technical effects as the foregoing embodiment due to the computer program loaded and run on the processor. The principle analysis can be referred to the related description of the method steps, which will not be repeated here.

[0164] The storage medium includes, for example, a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0165] The steps of the methods or algorithms described in connection with the embodiments disclosed herein can be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module can reside in random access memory (RAM), memory, read-only memory (ROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0166] In the description of the present specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" and the like means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Also, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. In addition, the person skilled in the art can combine and combine the different embodiments or examples described in the present specification and the features of the different embodiments or examples, without contradiction.

[0167] In addition, the term defining the features of "first", "second" can explicitly or implicitly include at least one of the features. In the description of the present application, the meaning of "a plurality of" is at least two, for example, two, three, etc., unless otherwise explicitly specified, only for the purpose of description, and cannot be understood as indicating or implying the relative importance or implicitly indicating the number of the indicated technical features.

[0168] Therefore, any process or method descriptions in the flow charts or otherwise described herein can be understood as representing code modules, segments, or portions of code that include one or more executable instructions for implementing the specified logical functions or processes, and the scope of preferred embodiments of the present application includes additional implementation in which the functions are performed in a different order, in substantially simultaneous fashion, or in reverse order, as will be understood by those skilled in the art of the embodiments to which the present application pertains.

[0169] The embodiments of the specific implementation are the preferred embodiments of the present application, and do not limit the protection scope of the present application, so that: any equivalent changes made according to the structure, shape, principle of the present application should be covered within the protection scope of the present application.

Claims

1. A fully closed-mode rolling method, characterized in that, include: The thickness measurement device with integrated temperature sensor acquires real-time thickness data of the plate to be rolled and real-time temperature data of the corresponding position. A temperature-sensitive deep learning model is pre-set, and the real-time temperature data, the real-time thickness data and historical rolling condition parameters are used as input to output a prediction matrix of the plate thickness change trend under different temperature evolution paths within a preset time period. Based on the plate thickness change trend prediction matrix, combined with the target plate thickness and the current mechanical characteristic parameters of the pressure roller, a dynamic compensation pressure roller gap sequence is calculated. The dynamic compensation pressure roller gap sequence includes the gap adjustment amount and adjustment sequence for several future control cycles. The dynamic compensation roller gap sequence is sent to the roller control system, which dynamically adjusts the roller gap, and synchronously collects the thickness feedback data and corresponding temperature feedback data of the plate after the roller adjustment and rolling are completed, and uploads the thickness feedback data and the temperature feedback data. The thickness measurement device with integrated temperature sensor integrates a spectrometer to collect spectral data of the plate surface in real time, and identifies the material type and mechanical property parameters of the currently rolled plate online through a preset material identification model; Based on the identified material type and mechanical property parameters, retrieve the corresponding material-specific prediction sub-model; If the board material is new and historical data is insufficient, transfer learning technology is used. Based on the prediction sub-model for similar materials, the parameters of the temperature-sensitive deep learning model are adjusted in combination with the mechanical performance parameters. When calculating the dynamic compensation roller gap sequence, a safety margin compensation term is added for new materials or materials with insufficient data; Real-time data on humidity, air pressure, and equipment vibration in the rolling environment are obtained through environmental sensors installed at the rolling site. The influence of humidity, air pressure, and equipment vibration data on temperature transfer and thickness variation during plate rolling was analyzed, and an environmental correction model was constructed. The humidity, air pressure, and equipment vibration data, along with the real-time temperature data, the real-time thickness data, the historical rolling condition parameters, the material type, and the mechanical property parameters, are input into the temperature-sensitive deep learning model or the material-specific prediction sub-model. The prediction matrix for the change trend of the plate thickness is corrected.

2. The rolling method in a fully closed mode according to claim 1, characterized in that, The method includes: When the temperature change rate of the board exceeds the preset value, the optimal stable temperature range is calculated by the temperature-sensitive deep learning model based on the material type, mechanical performance parameters, target thickness, humidity, air pressure and equipment vibration data of the current board. The temperature of the sheet metal is adjusted to the optimal stable temperature range by means of a heating or cooling device integrated into the rolling line, and the adjustment power of the heating or cooling device is dynamically matched according to the real-time thermal radiation data of the sheet metal. The temperature adjustment data after the plate temperature is stabilized, the material type and mechanical performance parameters, the humidity, air pressure and equipment vibration data are simultaneously input into the temperature-sensitive deep learning model to generate a low-fluctuation exclusive prediction matrix. The dynamic compensation roller gap sequence is calculated based on the low-fluctuation dedicated prediction matrix.

3. The rolling method in a fully closed mode according to claim 1, characterized in that, The method includes: When the rolled sheet is a dissimilar metal composite sheet, an ultrasonic flaw detector and a metallographic microscope probe are added to the thickness measurement device with the integrated temperature sensor to acquire the bonding data of the interface of the dissimilar metal composite sheet in real time. The bonding data includes bonding strength information and interface defect information. Spectral analysis was performed on each layer of the dissimilar metal composite plate, and the material type and mechanical property parameters of each layer of metal were accurately identified by a preset material identification model. At the same time, a comprehensive mechanical performance model of the dissimilar metal composite plate is established, taking into account the interaction between the metal layers, so as to more accurately describe the overall mechanical behavior of the dissimilar metal composite plate. Based on the material type, mechanical property parameters, and bonding data of each metal layer, a custom temperature-sensitive deep learning sub-model is created. During training, the temperature-sensitive deep learning sub-model focuses on the heat transfer and deformation coordination relationship between different metal layers to improve the accuracy of predicting the thickness change trend of the dissimilar metal composite plate. When calculating the dynamic compensation roller gap sequence, the target thickness of the plate and the mechanical characteristic parameters of the roller are considered, and the bonding data and the comprehensive mechanical property model of the dissimilar metal composite plate are combined. For the weak interface bonding areas, additional gap compensation is added to ensure uniform deformation and good bonding of each layer of metal in the dissimilar metal composite plate during the rolling process.

4. The rolling method in a fully closed mode according to claim 3, characterized in that, The method includes: By using multiple high-precision temperature sensors installed at the rolling site, real-time temperature change data of each layer of metal is obtained; Based on the thermal expansion coefficients of each metal layer, the thermal expansion at different locations is calculated, and a thermal expansion distribution map is generated. Based on the differences in thermal expansion of each metal layer in the thermal expansion distribution diagram, determine the metal layers that require temperature control and the direction of temperature control. The heating or cooling devices integrated into the rolling line allow for independent temperature control of each layer of metal. Real-time monitoring of changes in thermal expansion of each metal layer; adjustment of heating or cooling rates based on changes in thermal expansion. Acquire thermal expansion monitoring data for each metal layer; Obtain the temperature control parameters of the heating or cooling device; The thermal expansion monitoring data and the temperature control parameters are input into the temperature-sensitive deep learning sub-model to update the thickness change trend prediction matrix of the plate in real time, so as to more accurately reflect the influence of thermal expansion differences on the thickness and interface bonding of the dissimilar metal composite plate.

5. The rolling method in a fully closed mode according to claim 1, characterized in that, The method includes: Distributed fiber Bragg grating sensors are embedded inside the roller body and bearing housing to collect real-time temperature distribution data of the roller in the circumferential and axial directions, and obtain the temperature gradient of the roller surface and the temperature rise rate of the bearing housing. By combining the thermal expansion coefficient of the pressure roller material, a prediction model for the thermal deformation of the pressure roller is constructed. Based on the heat conduction equation and the expansion formula, the gap deviation caused by the radial expansion of the pressure roller due to temperature changes is calculated. The temperature distribution data, the gap deviation, the real-time thickness data, and the real-time temperature data are all input into the temperature-sensitive deep learning model to generate a thermally compensated dynamic gap sequence. The thickness feedback data of the rolled plate is collected by the thickness measuring device. If the thickness deviation exceeds the preset range within a continuous preset period and the deviation direction is consistent with the predicted direction of the heat deformation of the pressure roller, the actual temperature of the pressure roller, the actual gap between the pressure roller and the plate, and the actual thickness of the plate are extracted at the current moment. The pressure roller heat deformation weight parameters of the temperature-sensitive deep learning model are updated to achieve adaptive optimization of the compensation strategy.

6. A fully enclosed rolling mill, characterized in that, include: A memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and executed the fully closed-mode rolling method as described in any one of claims 1-5.

7. A storage medium, characterized in that, The computer program is stored that can be loaded by a processor and executed in the fully closed rolling mode as described in any one of claims 1-5.

Images