A multi-dimensional monitoring and control system for the closing accuracy of a roller ring mill
By using a full-coverage monitoring network and thermo-coupled calculations, the temperature anomalies and vibration effects of the ring rolling mill are identified in real time. A multi-dimensional monitoring and control system is constructed, which solves the problems of decreased accuracy and defects in the closed-loop control of the ring rolling mill and realizes accuracy prediction and dynamic regulation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- 江苏拢研机械有限公司
- Filing Date
- 2026-01-15
- Publication Date
- 2026-06-02
AI Technical Summary
Existing ring rolling mills suffer from decreased accuracy and defects due to temperature anomalies and vibration during the closed-loop control process. They lack a quantitative description of the intrinsic coupling relationship between vibration effects and the closed-loop state, making it difficult to accurately identify phase transition critical points and temperature anomalies.
By employing a full-coverage monitoring network and thermo-coupling calculations, and using an infrared thermal imager to monitor temperature anomalies in real time, a multi-dimensional monitoring and control system is constructed by combining vibration-closure influence strategies and differential constraint equations to achieve closure accuracy prediction and dynamic parameter control.
It enables real-time identification of phase transition critical points and early warning of temperature anomalies, reduces the risk of process defects, alleviates dimensional deviations and surface unevenness caused by vibration, and ensures closure accuracy and process stability.
Smart Images
Figure CN122125147A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of quality monitoring technology for ring rolling mills, specifically a multi-dimensional monitoring and control system for the closing accuracy of a ring rolling mill. Background Technology
[0002] In recent years, with the continuous improvement of the level of intelligence and automation in the manufacturing industry, the requirements for metal processing equipment have become increasingly higher. As a key piece of equipment, the closed-loop control process of the ring rolling mill directly affects the structural integrity and service life of the product. However, in actual production, the closed-loop control of the ring rolling mill faces many technical challenges and constraints, and these problems urgently need further exploration and solutions in both theory and practice.
[0003] During the ring forging closure process, metallic materials typically undergo phase transformations, accompanied by the release or absorption of latent heat, resulting in plateaus or inflection points in the temperature field. These temperature anomalies not only affect the mechanical state during closure but can also induce localized stress concentrations, ultimately leading to closure errors or defects. Simultaneously, equipment vibration can cause not only localized dimensional deviations but also uneven closure surfaces, stress concentrations, and structural fatigue, resulting in decreased accuracy or long-term drift, thus affecting the accuracy of monitoring data. Current descriptions of vibration impacts largely rely on empirical data and traditional detection methods, lacking the ability to quantify the intrinsic coupling relationship between vibration effects and the ring rolling mill's closure state. Therefore, how to rapidly and accurately identify phase transformation critical points and temperature anomalies from large amounts of real-time infrared thermal imaging data, and accurately capture and quantitatively describe the impact of vibration on the accuracy of prediction models and closure precision, has become an urgent problem to be solved.
[0004] Therefore, the present invention provides a multi-dimensional monitoring and control system for the closing accuracy of a ring rolling mill. Summary of the Invention
[0005] The purpose of this invention is to provide a multi-dimensional monitoring and control system for the closing accuracy of a ring rolling mill, so as to solve the existing problems mentioned in the background art.
[0006] To achieve the above objectives, the present invention provides the following technical solution: a multi-dimensional monitoring and control system for the closing accuracy of a ring rolling mill, comprising: The data acquisition module arranges temperature, pressure, displacement, and laser ranging sensors on the contact surface of the ring roll to form a full-coverage monitoring network, and divides the contact surface of the roll into uniform zones. The quality monitoring module obtains the thermo-mechanical coupling coefficient through a thermal stress coupling calculation strategy; The closure prediction module is used to establish a closure state prediction model, equipped with a vibration-closure effect strategy, taking the thermo-coupling coefficient and stress distribution as input data, and outputting the closure accuracy prediction value. The dynamic parameter control module is used to extract the predicted closure accuracy value and automatically adjust the control parameters of the ring rolling mill.
[0007] A further improvement of this invention is that the quality monitoring module includes a temperature anomaly monitoring unit, a stress distribution calculation unit, and a thermo-mechanical coupling coefficient calculation unit. The temperature anomaly monitoring unit is used to collect data on the contact surface area of the ring roll in real time using an infrared thermal imager, and to capture the influence of heat capacity and latent heat effect during the phase transformation process by utilizing temperature change characteristics, thereby obtaining a temperature anomaly index. The stress distribution calculation unit is used to calculate the distribution of internal residual stress by constructing a thermo-mechanical coupling model based on the measured temperature field and deformation field data, and to calculate the average residual stress in the contact surface area of the roll to obtain a stress influence coefficient. The thermo-mechanical coupling coefficient calculation unit is used to obtain a thermo-mechanical coupling coefficient by weighted summation of the stress influence coefficient and the temperature anomaly index.
[0008] A further improvement of this invention is that the temperature anomaly monitoring unit first converts radiation intensity into temperature T using the Stefan-Boltzmann law, calculates the rate of temperature change over time using the first derivative of temperature, and captures the inflection point of temperature change using the second derivative of temperature. When a plateau phenomenon occurs in the first derivative, the unit searches for the time point t corresponding to the moment when the second derivative of temperature is less than the set temperature threshold, and records it as the critical temperature moment of the metal phase transition. Then, the corresponding temperature T(t) is extracted to represent the phase transition temperature. The deviation ΔT(t) between the real-time temperature and the phase transition temperature is calculated, and the temperature anomaly deviation is output by comparing the temperature derivative of ΔT(t) with that of the ideal model Tid(t), and a temperature anomaly index is also output.
[0009] A further improvement of this invention is that the specific construction process of the thermo-mechanical coupling model includes: calculating the thermal strain at each point through the temperature field; in the elastic stage, using Hooke's law to obtain the stress field; setting the initial residual stress field distribution; using the finite element method to calculate the stress field and obtain the predicted deformation field; calculating the deformation error function Ede and its gradient information through the actual measured deformation field and the predicted deformation field; adjusting the assumed residual stress field through gradient descent; and repeating the process of predicting the deformation field and calculating the deformation error function Ede until Ede converges to a preset tolerance.
[0010] A further improvement of this invention lies in that the closure prediction module constructs a material-mass correlation model based on historical data and real-time monitoring data, including an input layer, a feature extraction layer, a training layer, and a vibration constraint layer. The input layer includes a thermal field, stress distribution, and vibration modes. The feature extraction layer extracts spatial features through a 3D convolutional neural network and captures long-range dependencies through a multi-head attention mechanism. The training layer includes historical thermal fields, stress distributions, vibration modes, and their corresponding closure accuracies. The vibration constraint layer defines a loss function by incorporating a vibration-closure influence strategy. ,in, This represents the predicted closure accuracy. This represents the actual value of the closure accuracy, and vie represents the vibration-induced behavior; the output is the predicted value of the closure accuracy.
[0011] A further improvement of this invention is that the material-mass correlation model further includes a vibration weight constraint layer. The vibration influence behavior weight of the vibration constraint layer is adjusted according to the deviation of the closure accuracy prediction value, and the calculation formula is as follows: ,in, This indicates the preset initial vibration influence behavior weight. This represents the correction factor.
[0012] A further improvement of this invention lies in that the vibration-closure effect strategy describes the influence of vibration on closure accuracy by introducing differential constraint equations; firstly, the residual stress field in each region of the roll contact surface is converted into equivalent stress excitation force using the stress gradient method. The integral of the temperature anomaly exponential field of the roll contact surface region is used as the thermo-equivalent excitation. By superimposing the individual components, coupled excitation constraints are obtained, and thus the vibration influence behavior is derived. Where m represents mass, c represents damping coefficient, k represents stiffness coefficient, and u represents the monitored displacement of the steel plate.
[0013] A further improvement of the present invention is that the dynamic parameter control module includes extracting the deviation between the predicted value of the closing accuracy output by the closing prediction module and the target value of the closing accuracy, generating a roll pressure adjustment command through PID, and when the closing accuracy after three consecutive adjustments is still greater than the target value of the closing accuracy, running a multi-objective optimization model to finely control the overall rolling pressure, balance size deviation, pressure offset state, residual stress and pressure of each roll.
[0014] A further improvement of this invention is that the multi-objective optimization model is achieved through... To achieve, among which, Indicates the stress influence coefficient. This represents the average actual dimensional deviation of each region, where P represents the actual applied overall pressure. The target is the overall pressure, and the pressure distribution of each region is output through a multi-roll pressure distribution strategy.
[0015] A further improvement of this invention is that the multi-roll pressure distribution strategy uses displacement sensors to monitor the actual dimensional deviation of the corresponding area of each roll and distributes the pressure to each area.
[0016] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention first captures the heat capacity and latent heat effect in the metal phase transformation process through the analysis of the first and second derivatives of temperature, and realizes the real-time identification of the phase transformation critical point and temperature anomaly, so that the system can give early warning of local temperature anomalies and reduce the risk of process defects caused by temperature anomalies. 2. Secondly, by using the vibration-closure influence strategy and differential constraint equations, the residual stress is converted into an equivalent excitation force, and implicit learning is achieved by combining thermal excitation. This compensates for the adverse effects of vibration on the closed state prediction model in real time, thereby compensating for the adverse effects of vibration on the closure accuracy. While ensuring the accuracy of the closure prediction, it effectively alleviates the problems of dimensional deviation and surface unevenness caused by vibration. 3. By adopting PID coarse adjustment to quickly respond to the closing accuracy deviation, and combining it with a multi-objective optimization model to finely control the roll pressure, pressure distribution and other process parameters, closed-loop control and dynamic parameter adjustment are realized. This scheme can ensure the overall process stability and closing quality while dealing with sudden disturbances. Attached Figure Description
[0017] Figure 1 This is a framework diagram of a multi-dimensional monitoring and control system for the closing accuracy of a ring rolling mill according to the present invention; Figure 2 This is a flowchart of the closure prediction module of a multi-dimensional monitoring and control system for the closure accuracy of a ring rolling mill according to the present invention; Figure 3 This is a flowchart of the dynamic parameter control module of a multi-dimensional monitoring and control system for the closure accuracy of a ring rolling mill according to the present invention. Detailed Implementation
[0018] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations thereof. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.
[0019] The term "and / or" simply describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. Additionally, the character " / " generally indicates that the preceding and following related objects have an "or" relationship.
[0020] Example 1 Figure 1 This embodiment illustrates a framework diagram of a multi-dimensional monitoring and control system for the closure accuracy of a ring rolling mill, including: The data acquisition module arranges temperature, pressure, displacement, and laser ranging sensors on the contact surface of the ring roll to form a full-coverage monitoring network, and divides the contact surface of the roll into uniform zones. Temperature monitoring combines contact and non-contact methods, covering both the material surface and internal thermal field; pressure sensors are embedded in the roll bearing housing to capture dynamic load changes in real time; displacement sensors use a combination of laser and mechanical probes to track the deformation trajectory of the ring component. The quality monitoring module obtains the thermo-mechanical coupling coefficient through a thermal stress coupling calculation strategy; The quality monitoring module includes a temperature anomaly monitoring unit, a stress distribution calculation unit, and a thermo-coupling coefficient calculation unit. The temperature anomaly monitoring unit is used to collect data of the contact surface area of the ring roll in real time using an infrared thermal imager, and to capture the influence of heat capacity and latent heat effect during the phase transformation process by utilizing temperature change characteristics to obtain a temperature anomaly index. The temperature anomaly monitoring unit first converts radiation intensity into temperature T using the Stefan-Boltzmann law, and then calculates the rate of temperature change over time using the first derivative of temperature. The first derivative reflects the rate of temperature rise or fall. During metal phase transitions, due to latent heat, the temperature rise often slows down or even stagnates, leading to a significant change in the first derivative. The inflection point of temperature change is captured through the second derivative. When a metal transitions from one phase to another, the temperature curve shows an inflection point due to the absorption or release of latent heat. At this time, the second derivative approaches zero and changes sign. By obtaining the maximum value of temperature change within time tti, when the maximum value is less than a set plateau threshold, it is judged as a plateau phenomenon, indicating that the metal may be undergoing a phase transition. When the first derivative shows a plateau phenomenon, the time point t corresponding to the moment when the second derivative of temperature is less than the set temperature threshold is searched and recorded as the critical temperature moment of the metal phase transition. Then, the corresponding temperature T(t) is extracted to represent the phase transition temperature. The deviation ΔT(t) between the real-time temperature and the phase transition temperature is calculated. The temperature anomaly deviation is output by comparing ΔT(t) with the temperature derivative of the ideal model Tid(t), and a temperature anomaly index is also output. The formula for calculating the temperature anomaly index is as follows: ; The stress distribution calculation unit is used to calculate the distribution of internal residual stress by constructing a thermo-mechanical coupling model based on the measured temperature field and deformation field data, and to calculate the average residual stress in the contact surface area of the roll of the workpiece to obtain the stress influence coefficient; it establishes a direct correlation between surface strain and residual stress, providing input for vibration prediction; the specific construction process of the thermo-mechanical coupling model includes: The thermal strain at each point is calculated by the temperature field. In the elastic stage, Hooke's law is used to obtain the stress field. The initial residual stress field distribution is set. The stress field is calculated using the finite element method to obtain the predicted deformation field. Through actual measured deformation field With the predicted deformation field Calculate the deformation error function Ede and its gradient information. The assumed residual stress field is adjusted by gradient descent, and the process of predicting the deformation field and calculating the deformation error function Ede is repeated until Ede converges to the preset tolerance.
[0021] The thermo-coupling coefficient calculation unit is used to obtain the thermo-coupling coefficient by weighted summation of the stress influence coefficient and the temperature anomaly index.
[0022] Example 2 Based on the inventive concept of Embodiment 1, this embodiment proposes a closure prediction module for establishing a closure state prediction model, equipped with a vibration-closure influence strategy, taking the thermo-coupling coefficient and stress distribution as input data, and outputting a closure accuracy prediction value. Figure 2 The flowchart of the closure prediction module of the multi-dimensional monitoring and control system for the closure accuracy of a ring rolling mill of the present invention is shown, which specifically includes: The closure prediction module constructs a material-mass correlation model based on historical data and real-time monitoring data, including an input layer, a feature extraction layer, a training layer, and a vibration constraint layer. The input layer includes thermal field, stress distribution, and vibration modes. Since vibration affects mass during the closing process, and residual stress affects vibration, a vibration label is added to the input layer to calculate the vibration state during closing, and then combined with the metal material state to predict the possible dimensional deviations during the closing process. The feature extraction layer extracts spatial features through a 3D convolutional neural network and captures long-range dependencies through a multi-head attention mechanism. The training layer includes historical thermal fields, stress distributions, vibration modes, and their corresponding closure accuracy; the evaluation index for closure accuracy includes diameter deviation. Ellipticity error Surface flatness , The surface height distribution is represented by the above parameters, which are then standardized and weighted to obtain the closure accuracy.
[0023] The vibration constraint layer defines a loss function by incorporating a vibration-closing effect strategy. ,in, This represents the predicted closure accuracy. Here, represents the actual value of the closure accuracy, and vie represents the vibration-induced behavior. By combining data-driven approaches with physical mechanism constraints, the model can accurately predict the closure accuracy while conforming to vibration laws, thus implicitly learning the vibration effects and achieving compensation. This is achieved by minimizing... This forces the model parameters to be adjusted so that the predicted vibration response meets the physical laws, and outputs the predicted value of the closure accuracy.
[0024] The material-mass correlation model also includes a vibration weight constraint layer. The vibration influence behavior weights of the vibration constraint layer are adjusted according to the deviation of the closure accuracy prediction value. The calculation formula is as follows: ,in, This indicates the preset initial vibration influence behavior weight. This represents the correction factor.
[0025] The vibration-closure effect strategy describes the impact of vibration on closure accuracy by introducing differential constraint equations. First, the residual stress field in each region of the roll contact surface is converted into an equivalent stress excitation force using the stress gradient method. The integral of the temperature anomaly exponential field of the roll contact surface region is used as the thermo-equivalent excitation. The coupled excitation constraints are obtained by superimposing the individual components. This leads to the understanding of the influence of vibration on behavior. Where m represents mass, c represents damping coefficient, k represents stiffness coefficient, and u represents the monitored steel plate displacement. These fields represent dividing the closed region into n regions on average according to the contact surface of the rolls.
[0026] Predictive models can provide lead time for control methods, allowing for time to adjust parameters and avoiding post-hoc remediation.
[0027] Example 3 Based on the inventive concepts of Embodiments 1 and 2, this embodiment proposes a dynamic parameter control module for extracting the predicted closure accuracy value and automatically adjusting the control parameters of the ring rolling mill. Figure 3 The flowchart of the dynamic parameter control module of the multi-dimensional monitoring and control system for the closure accuracy of a ring rolling mill of the present invention is shown, which specifically includes: The deviation between the predicted closure accuracy value output by the closure prediction module and the target closure accuracy value is extracted, and a roll pressure adjustment command is generated using a PID controller. The PID algorithm has a simple structure and fast response speed, making it suitable for handling sudden disturbances (such as the impact at the moment of roll bite). Its linear combination form can take into account the current error, historical cumulative error, and future trend, achieving rapid coarse adjustment. When the closure accuracy after three consecutive adjustments is still greater than the target closure accuracy value, it indicates that the coarse adjustment cannot meet the closure error requirements. Therefore, a multi-objective optimization model is run to finely control the overall rolling pressure, balance dimensional deviation, pressure offset state, residual stress, and the pressure of each roll.
[0028] The multi-objective optimization model is implemented through the following formula: ; in, Indicates the stress influence coefficient. This represents the average actual dimensional deviation of each region, where P represents the actual applied overall pressure. Overall pressure on the target, due to This represents a result variable, not a directly controllable parameter. Therefore, the parameters actually adjusted by the fine-tuning layer are directly controlled process parameters. At this point, the overall rolling pressure is output through the multi-roll pressure distribution strategy. and pressure distribution in different regions .
[0029] The multi-roll pressure distribution strategy uses displacement sensors to monitor the actual dimensional deviation of each roll's corresponding area and distributes the pressure to each area. The pressure of the i-th area is represented as... Using inverse weights Distribute pressure; the greater the deformation, the smaller the pressure, thus suppressing local overload; at the same time, normalize the denominator. Ensure the total pressure remains constant and complies with the equipment's load-bearing limits.
[0030] This invention ensures process stability through coarse adjustment using a PID algorithm and optimizes global performance through fine adjustment. The two work together to improve closure accuracy. If only coarse adjustment is relied upon, complex coupling problems may not be solved. If only fine adjustment is relied upon using a multi-objective optimization model, the response speed will be insufficient. Therefore, by combining coarse and fine adjustment with control, fast response can be achieved while ensuring closure accuracy.
[0031] The threshold and weight settings can be based on the default settings of this invention, or they can be set by the operator.
[0032] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0033] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0034] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0035] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0036] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.
Claims
1. A multi-dimensional monitoring and control system for the closing accuracy of a ring rolling mill, characterized in that: include: The data acquisition module arranges temperature, pressure, displacement, and laser ranging sensors on the contact surface of the ring roll to form a full-coverage monitoring network, and divides the contact surface of the roll into uniform zones. The quality monitoring module obtains the thermo-mechanical coupling coefficient through a thermal stress coupling calculation strategy; The closure prediction module is used to establish a closure state prediction model, equipped with a vibration-closure effect strategy, taking the thermo-coupling coefficient and stress distribution as input data, and outputting the closure accuracy prediction value. The dynamic parameter control module is used to extract the predicted closure accuracy value and automatically adjust the control parameters of the ring rolling mill.
2. The multi-dimensional monitoring and control system for the closing accuracy of a ring rolling mill according to claim 1, characterized in that: The quality monitoring module includes a temperature anomaly monitoring unit, a stress distribution calculation unit, and a thermo-coupling coefficient calculation unit. The temperature anomaly monitoring unit is used to collect data of the contact surface area of the ring roll in real time using an infrared thermal imager, and to capture the influence of heat capacity and latent heat effect during the phase transformation process by utilizing temperature change characteristics to obtain a temperature anomaly index. The stress distribution calculation unit is used to calculate the distribution of internal residual stress by constructing a thermo-mechanical coupling model based on the measured temperature field and deformation field data, and to calculate the average residual stress in the contact surface area of the roll of the workpiece to obtain the stress influence coefficient; the thermo-mechanical coupling coefficient calculation unit is used to obtain the thermo-mechanical coupling coefficient by weighted summation of the stress influence coefficient and the temperature anomaly index.
3. The multi-dimensional monitoring and control system for the closing accuracy of a ring rolling mill according to claim 2, characterized in that: The temperature anomaly monitoring unit first converts radiation intensity into temperature T using the Stefan-Boltzmann law, calculates the rate of temperature change over time using the first derivative of temperature, and captures the inflection point of temperature change using the second derivative of temperature. When a plateau phenomenon occurs in the first derivative, it searches for the time point t corresponding to the moment when the second derivative of temperature is less than the set temperature threshold, and records it as the critical temperature moment of the metal phase transition. Then, it extracts the corresponding temperature T(t) to represent the phase transition temperature. It calculates the deviation ΔT(t) between the real-time temperature and the phase transition temperature, outputs the temperature anomaly deviation degree by comparing the temperature derivative of ΔT(t) with that of the ideal model Tid(t), and outputs the temperature anomaly index.
4. The multi-dimensional monitoring and control system for the closing accuracy of a ring rolling mill according to claim 2, characterized in that: The specific construction process of the thermo-mechanical coupling model includes: calculating the thermal strain at each point through the temperature field; in the elastic stage, using Hooke's law to obtain the stress field; setting the initial residual stress field distribution; using the finite element method to calculate the stress field and obtain the predicted deformation field; calculating the deformation error function Ede and its gradient information through the actual measured deformation field and the predicted deformation field; adjusting the assumed residual stress field through gradient descent; and repeating the process of predicting the deformation field and calculating the deformation error function Ede until Ede converges to within the preset tolerance.
5. The multi-dimensional monitoring and control system for the closing accuracy of a ring rolling mill according to claim 1, characterized in that: The closure prediction module constructs a material-quality correlation model based on historical and real-time monitoring data, including an input layer, a feature extraction layer, a training layer, and a vibration constraint layer. The input layer includes thermal field, stress distribution, and vibration modes. The feature extraction layer extracts spatial features through a 3D convolutional neural network and captures long-range dependencies using a multi-head attention mechanism. The training layer includes historical thermal field, stress distribution, vibration modes, and their corresponding closure accuracy. The vibration constraint layer defines a loss function using a vibration-closure influence strategy. ,in, This represents the predicted closure accuracy. This represents the actual value of the closure accuracy, and vie represents the vibration-induced behavior; the output is the predicted value of the closure accuracy.
6. A multi-dimensional monitoring and control system for the closing accuracy of a ring rolling mill according to claim 5, characterized in that: The material-mass correlation model also includes a vibration weight constraint layer. The vibration influence coefficient weight of the vibration constraint layer is adjusted according to the deviation of the predicted closure accuracy value. The calculation formula is as follows: ,in, This indicates the preset initial vibration influence coefficient weight. This represents the correction factor.
7. A multi-dimensional monitoring and control system for the closing accuracy of a ring rolling mill according to claim 5, characterized in that: The vibration-closure effect strategy describes the impact of vibration on closure accuracy by introducing differential constraint equations. First, the residual stress field in each region of the roll contact surface is converted into an equivalent stress excitation force using the stress gradient method. The integral of the temperature anomaly exponential field of the roll contact surface region is used as the thermo-equivalent excitation. By superimposing the individual components, coupled excitation constraints are obtained, and thus the vibration influence behavior is derived. Where m represents mass, c represents damping coefficient, k represents stiffness coefficient, and u represents the monitored displacement of the steel plate.
8. The multi-dimensional monitoring and control system for the closing accuracy of a ring rolling mill according to claim 1, characterized in that: The dynamic parameter control module includes extracting the deviation between the predicted value of the closing accuracy output by the closing prediction module and the target value of the closing accuracy, generating roll pressure adjustment commands through PID, and running a multi-objective optimization model when the closing accuracy after three consecutive adjustments is still greater than the target value of the closing accuracy, thereby finely controlling the overall rolling pressure, balance size deviation, pressure offset state, residual stress and pressure of each roll.
9. A multi-dimensional monitoring and control system for the closing accuracy of a ring rolling mill according to claim 8, characterized in that: The multi-objective optimization model is achieved through... To achieve, among which, Indicates the stress influence coefficient. This represents the average actual dimensional deviation of each region, where P represents the actual applied overall pressure. The target is the overall pressure, and the pressure distribution of each region is output through a multi-roll pressure distribution strategy.
10. A multi-dimensional monitoring and control system for the closing accuracy of a ring rolling mill according to claim 9, characterized in that: The multi-roll pressure distribution strategy uses displacement sensors to monitor the actual dimensional deviation of the corresponding area of each roll and distributes the pressure to each area.