Intelligent temperature control system and method for non-oriented silicon steel continuous annealing process

Abstract

The invention discloses an intelligent temperature control system and method for a non-oriented silicon steel continuous annealing process, and relates to the technical field of non-oriented silicon steel temperature control, the intelligent temperature control system comprises a multi-source sensing fusion acquisition module, data are acquired through multi-modal sensing and edge calculation, a depth space-time modeling module is used for modeling prediction through LSTM and GNN, and the temperature of the non-oriented silicon steel continuous annealing process is obtained. The intelligent cooperative regulation and control module combines reinforcement learning with MPC temperature control and further comprises a monitoring feedback module, a multi-dimensional early warning module and a data insight optimization module, and all the modules work cooperatively. According to the invention, the temperature control precision is improved, the defective rate is reduced, the production efficiency is improved, the energy consumption is reduced, the quality is scientifically monitored, the parameters are adaptively adjusted, the remote operation is realized, and the continuous annealing temperature control of the non-oriented silicon steel is comprehensively optimized.

Description

Technical Field

[0001] The present invention relates to the technical field of temperature control of non-oriented silicon steel, and particularly to an intelligent temperature control system and method for the continuous annealing process of non-oriented silicon steel. Background Art

[0002] As an important soft magnetic material in the power industry, non-oriented silicon steel is widely used in various equipment such as motors and transformers. Its continuous annealing process plays a decisive role in product quality, and temperature control during the annealing process is the core link of the process. Traditional temperature control means for the continuous annealing of non-oriented silicon steel mostly rely on manual experience and simple automation control. In this mode, operators set the temperature parameters of the annealing furnace based on accumulated experience. However, manual judgment is not only inefficient but also easily affected by subjective factors, making it difficult to accurately respond to complex and changing production conditions.

[0003] With the continuous expansion of the production scale of non-oriented silicon steel and the increasingly stringent requirements for product quality, the limitations of traditional temperature control systems have become more prominent. Simple automation control can maintain temperature stability to a certain extent, but it cannot fully consider the comprehensive effects of various factors such as the chemical composition differences of silicon steel, the running speed and thickness changes of the strip on the annealing temperature. For example, there are slight fluctuations in the chemical composition of different batches of silicon steel, and traditional systems are difficult to adjust the temperature in a timely and accurate manner for these changes, resulting in poor product performance consistency and a high rejection rate.

[0004] In the context of the booming development of Industry 4.0 and intelligent manufacturing, the continuous annealing process of non-oriented silicon steel urgently needs intelligent upgrading. Although there are some temperature control systems on the market that attempt to introduce intelligent technologies, most of them are only superficially applied, such as simple data collection and basic temperature feedback regulation, lacking in-depth understanding and modeling analysis of complex physical phenomena during the annealing process. This makes the system unable to make rapid and effective responses in the face of emergencies or complex process requirements, severely restricting the further improvement of the production efficiency and product quality of non-oriented silicon steel. Therefore, developing a system that can comprehensively sense production information, accurately model and predict temperature changes, and achieve intelligent and efficient temperature control has become an urgent need to promote the development of the non-oriented silicon steel industry. Summary of the Invention

[0005] The intelligent temperature control system and method for the continuous annealing process of non-oriented silicon steel proposed by the present invention are to solve the problems mentioned in the above prior art.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] An intelligent temperature control system for the continuous annealing process of non-oriented silicon steel, comprising the following modules:

[0008] Multi-source perception fusion data acquisition module: It adopts a multi-modal sensor array including temperature, speed, thickness measurement, and monitoring of the microstructure of silicon steel. The temperature sensors are distributed by fiber Bragg gratings, the speed sensors and thickness gauges are self-calibrated, and the edge computing technology is used to preliminarily process the raw data;

[0009] Deep spatio-temporal modeling module: It constructs a prediction model based on the long short-term memory network (LSTM) and the graph neural network (GNN). The LSTM captures the dynamic changes in the continuous annealing of silicon steel, and the GNN learns the heat transfer correlation in the spatial dimension;

[0010] Intelligent collaborative temperature control module: It combines reinforcement learning and model predictive control (MPC), comprehensively considering the control objectives and system constraints, and improves the formula to adjust the power, where α, β, and γ are dynamically adjusted coefficients, is the predicted temperature at the future k-th moment, and N is the prediction horizon;

[0011] Panoramic real-time monitoring and feedback module: It deploys a distributed fiber optic sensing network and thermal imaging cameras to monitor the temperature of the annealing furnace and the silicon steel strip, and constructs a temperature monitoring system; The quality is evaluated by a multi-sensor fusion algorithm. When an anomaly occurs, the deviation is transmitted to the temperature control module through the feedback mechanism for closed-loop control, and the monitoring data is encrypted and stored in a distributed ledger database;

[0012] Multi-dimensional intelligent early warning module: It establishes an abnormal temperature early warning model of a multi-variable Bayesian network, constructs a probabilistic causal network, and uses the improved formula to calculate the early warning index, where ρ i is the weight of different abnormal events, P(A i |E) is the posterior probability of the abnormal event A i occurring under the current evidence E, and l is the number of abnormal events; When W exceeds the threshold, an audible and visual alarm is activated, information is pushed to the mobile terminal through the Internet of Things, and a fault diagnosis report is automatically generated;

[0013] Deep data insight and optimization module: It uses the technology combining deep autoencoders and generative adversarial networks to mine the monitoring and historical data, and uses association rule mining and clustering analysis to find the reasons affecting the annealing temperature and the quality of silicon steel, and adaptively optimizes according to the mining results.

[0014] Furthermore, it also includes an adaptive dynamic parameter adjustment module. Based on the production conditions and quality feedback, this module uses the genetic algorithm to adjust the LSTM and GNN network parameters in the deep spatio-temporal modeling module, as well as the reinforcement learning reward function and MPC prediction horizon parameters in the intelligent collaborative temperature control module, and uses the adaptive adjustment formula to optimize the model and control parameters, where θ old is the old parameter, and δ is the adaptive learning rate, is the gradient of the loss function with respect to the parameters.

[0015] Furthermore, it also includes a remote holographic monitoring and operation module. This module utilizes 5G communication technology and holographic projection technology to achieve remote monitoring and operation of the annealing furnace temperature and production process. Operators can view the actual situation inside the annealing furnace in the form of holographic images through remote terminal devices, support remote operation of functions such as adjusting the power of heating elements and starting and stopping equipment, and have functions for hierarchical management of operation permissions and recording operation logs.

[0016] Furthermore, the sensors in the multi-source perception fusion data acquisition module have self-diagnosis and self-repair functions. Through built-in fault detection algorithms, the working status of the sensors is monitored. When a fault is detected, standby sensors are activated, and microelectromechanical system (MEMS) technology and intelligent materials are used to automatically repair the faulty sensors. The multi-source data cross-validation technology is adopted to improve data reliability.

[0017] Furthermore, the input of the deep spatio-temporal modeling module covers the chemical composition of silicon steel, strip speed, thickness, and real-time temperature, as well as microstructure evolution data. By training on historical and real-time production data, an improved loss function is used to optimize the model parameters, where is the predicted temperature, is the true temperature, ω i is the weight of different data points, λ is the regularization coefficient, and θ j is the model parameter; Transfer learning technology is adopted to pre-train the model using data from other similar metal heat treatment processes, and then fine-tune the data for the continuous annealing of non-oriented silicon steel; Model compression technology is used to reduce the model storage and computational overhead.

[0018] Furthermore, the intelligent collaborative temperature control module introduces fuzzy adaptive PID control as an auxiliary control strategy. When uncertainties occur in the system, it automatically switches to fuzzy adaptive PID control, and based on the temperature deviation and the rate of change of the deviation, uses fuzzy rules to dynamically adjust the parameters of the PID controller and works in collaboration with reinforcement learning and MPC composite control.

[0019] Furthermore, the real-time monitoring and feedback module uses digital twin technology to construct a virtual model of the annealing furnace and the silicon steel production process. By synchronizing with the actual production data, it simulates the annealing process, predicts potential problems, and provides a virtual verification environment for optimizing control strategies.

[0020] Furthermore, a method for the intelligent temperature control system of the non-oriented silicon steel continuous annealing process includes the following steps:

[0021] Multi-source perception data acquisition step: Use a multi-modal sensor array to collect data on the temperature in each area of the annealing furnace, the speed, thickness, and microstructure evolution of the silicon steel strip. Initially process the data at the sensor nodes through edge computing, use fiber Bragg grating temperature sensing technology and self-calibrating measuring instruments for real-time acquisition, and adopt multi-source data cross-validation to improve data reliability;

[0022] Deep spatio-temporal modeling prediction step: Input the collected multi-source data into a joint prediction model based on LSTM and GNN, use an improved loss function for training, utilize transfer learning to accelerate model convergence, and combine model compression technology to improve online operation efficiency;

[0023] Intelligent collaborative temperature control step: Adopt a composite control strategy of reinforcement learning and MPC, combine fuzzy adaptive PID for auxiliary control, and adjust the power of the heating element using an improved power regulation formula based on the real-time temperature state and prediction results;

[0024] Panoramic real-time monitoring and feedback step: Through a distributed fiber optic sensing network, thermal imaging cameras, and multi-sensor data fusion, panoramically and real-time monitor the temperature of the annealing furnace and the quality parameters of the silicon steel. Once an anomaly is detected, it is fed back to the temperature control step, and at the same time, the data is encrypted and stored in a distributed ledger database;

[0025] Multi-dimensional intelligent warning step: Based on a multi-variable Bayesian network warning model, use an improved warning formula to calculate warning indicators. When the indicators exceed the threshold, immediately trigger an alarm and push warning information, and at the same time generate a fault diagnosis report;

[0026] Deep data insight and optimization step: Use deep autoencoders and GANs to deeply mine the data, combine association rule mining and clustering analysis to discover key factors and patterns, and adaptively optimize the model and control strategy.

[0027] Furthermore, it also includes an adaptive dynamic parameter adjustment step. This step is based on the real-time production conditions and quality feedback, uses a genetic algorithm to dynamically adjust the model and control parameters, and optimizes the parameters using an adaptive adjustment formula.

[0028] Furthermore, it also includes a remote holographic monitoring and operation step. This step uses 5G and holographic projection technology to achieve remote holographic monitoring and operation, with permission management and operation log recording functions.

[0029] Compared with the existing technologies, the beneficial effects of the present invention are:

[0030] In terms of temperature control accuracy, the multi-source perception fusion data acquisition module cooperates with the deep spatio-temporal modeling module to comprehensively capture production process information and accurately predict temperature changes. The intelligent collaborative temperature control module uses advanced control strategies to improve the temperature control accuracy in the annealing furnace to within ±1°C, greatly reducing the product performance differences caused by temperature fluctuations, effectively improving the consistency of product quality, and reducing the defective product rate.

[0031] The production efficiency has been greatly improved. Based on the rapid processing and intelligent decision-making of real-time data, the system can automatically optimize temperature control parameters according to the characteristics of silicon steel and process requirements, without frequent manual intervention, reducing production adjustment time. At the same time, the fault warning and rapid repair mechanism ensure the continuity of production, significantly reducing the number of production interruptions and increasing the overall production efficiency by more than 30%.

[0032] The cost control effect is obvious. Precise temperature control avoids energy waste caused by improper temperature, reducing energy consumption by about 20%. The decrease in the defective product rate reduces raw material waste and rework costs, significantly reducing the comprehensive production cost.

[0033] Quality monitoring and management are more scientific. The panoramic real-time monitoring and feedback module realizes the full-scale monitoring of the annealing process and product quality. The multi-dimensional intelligent warning module discovers potential problems in advance, providing strong support for quality control. The deep data insight and optimization module mines data value, continuously optimizes the production process, further improves product quality, and enhances the market competitiveness of products.

[0034] In addition, the remote holographic monitoring and operation module breaks geographical restrictions, facilitating operators to remotely manage production and improving management efficiency. The adaptive dynamic parameter adjustment module ensures that the system can flexibly adapt to different production conditions and always maintain the best operating state, providing a solid guarantee for the intelligent and efficient development of the non-oriented silicon steel continuous annealing process. Description of the Drawings

[0035] Figure 1 It is a schematic block diagram of the intelligent temperature control system for the non-oriented silicon steel continuous annealing process proposed by the present invention;

[0036] Figure 2 It is a schematic block diagram of the intelligent temperature control method for the non-oriented silicon steel continuous annealing process proposed by the present invention;

[0037] Figure 3 It is a column chart comparing the performance consistency of non-oriented silicon steel products in different batches of the intelligent temperature control system and method for the non-oriented silicon steel continuous annealing process proposed by the present invention;

[0038] Figure 4 It is a line chart showing the relationship between energy consumption and production efficiency of the intelligent temperature control method for the non-oriented silicon steel continuous annealing process proposed by the present invention. Detailed implementation manners

[0039] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0040] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation of the present invention.

[0041] In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of the described features. In the description of the present invention, "a plurality of" means two or more unless otherwise specifically defined. In addition, the terms "mounted", "connected" and "connected" should be understood in a broad sense. For example, it may be a fixed connection, a detachable connection or an integral connection; it may be a mechanical connection or an electrical connection; it may be directly connected or indirectly connected through an intermediate medium, and it may be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations. The present invention will be further described in detail below with reference to the drawings.

[0042] Refer to Figures 1 to 4 : An intelligent temperature control system for a non-oriented silicon steel continuous annealing process, including the following modules:

[0043] Multi-source perception fusion data acquisition module: A multi-modal sensor array is used for data acquisition. Among them, new fiber Bragg grating temperature sensors are distributedly installed at various key positions in the annealing furnace, such as the heating zone, the heat preservation zone, and the cooling zone, etc. These sensors utilize the sensitive characteristics of fiber Bragg gratings to temperature changes and can collect temperature data at an ultra-high frequency of 20 times per second, with an accuracy of up to ±0.1°C. At the same time, a speed sensor and a thickness gauge are respectively installed on the conveying path of the silicon steel strip to monitor the running speed and thickness of the strip in real time. The speed sensor adopts a high-precision optoelectronic encoder, and the thickness gauge utilizes the principle of laser ranging, and they both have self-calibration functions and automatically perform calibration once every 10 minutes to ensure the accuracy of measurement. The newly added silicon steel microstructure monitoring sensor based on spectral analysis can obtain the microstructure evolution data of silicon steel during the annealing process in real time by emitting a specific wavelength of spectrum on the surface of the silicon steel strip and analyzing the characteristics of the reflected spectrum. At the sensor node, edge computing technology is used to preliminarily process the collected raw data, such as removing noise, performing data normalization, etc., and then transmit the processed data to the central control system, which can reduce the data transmission volume and improve the real-time performance of data acquisition.

[0044] Deep spatio-temporal modeling module: A joint prediction model is constructed based on the long short-term memory network (LSTM) and the graph neural network (GNN). First, a large amount of historical production data is collected, including the chemical composition of silicon steel, strip speed, thickness, real-time temperature, and microstructure evolution data, etc. After cleaning and preprocessing these data, they are used as the training set and input into the model. The LSTM network is used to capture the dynamic change of temperature over time during the continuous annealing process of silicon steel. Through memory units and gating mechanisms, it can effectively handle the long-term dependencies in time series data. The GNN learns the heat transfer correlations in the spatial dimension according to the spatial topological relationships of each region in the annealing furnace and the running path of the silicon steel strip. The input layer of the model receives multi-source data. After calculation and feature extraction by the hidden layer, the predicted annealing temperature is output. During the training process,

[0045] An improved loss function is used to optimize the model parameters, where is the predicted temperature, is the true temperature, ω i is the weight of different data points, which is set according to the importance and reliability of the data, and λ is the regularization coefficient used to prevent the model from overfitting, and θ j are the model parameters. By continuously adjusting the model parameters, the value of the loss function is minimized, thereby improving the prediction accuracy and generalization ability of the model.

[0046] Intelligent Collaborative Temperature Control Module: Adopts a composite control strategy that combines reinforcement learning and model predictive control (MPC). The reinforcement learning agent continuously learns by trial and error in the annealing temperature control environment. Based on information such as the current temperature state, temperature deviation, and deviation change rate, it selects the optimal power adjustment action for the heating element to maximize the long-term cumulative reward (such as improving temperature control accuracy and meeting the silicon steel quality standard). MPC plans the power adjustment sequence of the heating element in advance according to the annealing temperature prediction model. While considering the control objective, it also takes into account the system constraints, such as the maximum power limit of the heating element and the maximum rate of temperature change. In the actual control process, an improved power adjustment formula is used to adjust the power of the heating element in real time, where α, β, and γ are dynamically adjusted coefficients, is the predicted temperature at the future k-th moment, and N is the prediction horizon, to achieve precise and efficient temperature control.

[0047] Panoramic Real-time Monitoring and Feedback Module: Deploys a distributed optical fiber sensing network and high-definition thermal imaging cameras to achieve panoramic real-time monitoring of the annealing furnace. The distributed optical fiber sensing network is used to monitor the temperature distribution of the annealing furnace wall and the gas inside the furnace. It senses temperature changes through the transmission of optical signals in the optical fiber and has the characteristics of being distributed and highly accurate. The high-definition thermal imaging camera captures the temperature field on the surface of the silicon steel strip in real time with a 360° view, and can intuitively display the temperature distribution of the silicon steel strip during the annealing process. At the same time, key quality parameters such as the magnetic permeability and iron loss of the silicon steel are detected online. Through the multi-sensor data fusion algorithm, the temperature data and quality parameter data are fused and processed to evaluate the quality status of the silicon steel in real time. Once temperature anomalies or quality parameter deviations from the standard are detected, the system immediately feeds back the deviation information to the intelligent collaborative temperature control module to achieve closed-loop control. The monitoring data is stored in the distributed ledger database in encrypted form to ensure the immutability and traceability of the data. The distributed ledger database adopts blockchain technology, disperses the data storage on multiple nodes, each node stores a complete data copy, and ensures data consistency through the consensus mechanism.

[0048] Multi-dimensional Intelligent Early Warning Module: Establishes an abnormal temperature early warning model based on a multi-variable Bayesian network, comprehensively considering multi-dimensional information such as the temperature inside the annealing furnace, the operating parameters of the silicon steel strip, the microstructure changes, and the quality parameters. By learning from historical fault data and normal operation data, a probabilistic causal relationship network between variables is constructed. An improved early warning formula is used to calculate the early warning index W, where ρ i is the weight of different abnormal events, and P(A i |E) is the abnormal event A under the current evidence E iThe posterior probability of occurrence, where l is the number of abnormal events. When W exceeds the set threshold, the system immediately activates the audible and visual alarm, and pushes the early warning information to the mobile terminals of relevant personnel through the Internet of Things. At the same time, it automatically generates a fault diagnosis report to provide a basis for timely handling.

[0049] Deep Data Insights and Optimization Module: Utilize the technology that combines a deep autoencoder and a generative adversarial network (GAN) to deeply mine the monitoring data and historical data. The deep autoencoder performs dimensionality reduction on high-dimensional and complex data to extract the potential feature patterns of the data. It maps the input data to a low-dimensional feature space through the encoder, and then reconstructs the data in the feature space into the original data through the decoder, learning the feature representation of the data by minimizing the reconstruction error. GAN consists of a generator and a discriminator. The generator attempts to generate simulated annealing data, and the discriminator is responsible for distinguishing between real data and simulated data. Through the adversarial training of the two, the model can better learn the distribution characteristics of the data. Use data mining algorithms such as association rule mining and clustering analysis to discover the key factors and hidden laws that affect the annealing temperature and the quality of silicon steel. For example, through association rule mining, the association relationship between the chemical composition of silicon steel and the annealing temperature can be discovered, and through clustering analysis, silicon steel products of different qualities can be classified to find the common factors affecting product quality. According to the mining results, adaptively optimize the model structure and parameters of the deep spatio-temporal modeling module, as well as the control strategies and parameters of the intelligent collaborative temperature control module. For example, if it is found that a certain chemical composition has a greater impact on the annealing temperature, the weight of this chemical composition in the deep spatio-temporal modeling module can be adjusted; if it is found that a certain control strategy has poor effects under specific working conditions, the control strategy of the intelligent collaborative temperature control module can be optimized.

[0050] In the present invention, there is also an adaptive dynamic parameter adjustment module. This module, based on the real-time production conditions and the feedback of the quality of silicon steel, uses a genetic algorithm to dynamically adjust the LSTM and GNN network structures and parameters in the deep spatio-temporal modeling module, as well as key parameters such as the reinforcement learning reward function and the MPC prediction horizon in the intelligent collaborative temperature control module. Through the analysis of the characteristics of silicon steel in different production batches (such as chemical composition fluctuations, differences in initial microstructures) and process requirements (such as different magnetic property targets), use the adaptive adjustment formula to optimize the model and control parameters, where θ old is the old parameter, δ is the adaptive learning rate, is the gradient of the loss function with respect to the parameter, ensuring that the system always maintains the optimal control performance.

[0051] In the present invention, there is also a remote holographic monitoring and operation module. With the help of 5G communication technology and holographic projection technology, it realizes remote holographic monitoring and operation of the annealing furnace temperature and production process. High-definition cameras, sensors and other devices are deployed on-site of the annealing furnace to collect information such as the temperature inside the annealing furnace and the running state of the silicon steel strip in real time, and transmit this information to the remote terminal device through the 5G network. The remote terminal device uses holographic projection technology to present the collected information in the form of a holographic image, and the operator can view the real-time situation inside the annealing furnace from all directions and multiple angles. At the same time, the remote terminal device supports the operator to remotely operate functions such as power adjustment of heating elements and start / stop of equipment. To ensure the safety and traceability of remote operations, the system has a hierarchical management of operation permissions and a real-time recording function of operation logs. Only authorized personnel can perform remote operations, and each operation will be recorded in the operation log, including information such as operation time, operator, and operation content.

[0052] In the present invention, the sensors in the multi-source perception fusion data acquisition module have self-diagnosis and self-repair functions. Through the built-in fault detection algorithm, the working state of the sensors is monitored in real time. Once a fault is detected, the standby sensors are immediately activated, and microelectromechanical system (MEMS) technology and intelligent materials are used to automatically repair the faulty sensors to ensure uninterrupted data acquisition. At the same time, multi-source data cross-validation technology is adopted to improve data reliability.

[0053] In the present invention, the deep spatio-temporal modeling module adopts transfer learning technology. It pre-trains the model using data from other similar metal heat treatment processes, and then fine-tunes it for non-oriented silicon steel continuous annealing data to accelerate model convergence and improve the performance of the model under small-sample data. At the same time, model compression technologies such as pruning and quantization are used to reduce model storage and computational overhead and improve the online operation efficiency of the model.

[0054] In the present invention, the intelligent collaborative temperature control module introduces fuzzy adaptive PID control as an auxiliary control strategy. When the temperature changes violently or the system uncertainty is large, it automatically switches to fuzzy adaptive PID control. According to the temperature deviation and the rate of change of the deviation, the parameters of the PID controller are dynamically adjusted using fuzzy rules, and it works in cooperation with the reinforcement learning-MPC composite control to further improve the temperature control stability and robustness.

[0055] In the present invention, the panoramic real-time monitoring and feedback module uses digital twin technology to build a virtual model of the annealing furnace and the silicon steel production process. By synchronizing with the actual production data in real time, it simulates the annealing process, predicts potential problems in advance, and provides a virtual verification environment for optimizing control strategies to achieve efficient monitoring and feedback control combining the virtual and the real.

[0056] In the present invention, the following steps are included:

[0057] Multi-source perception data acquisition step: Use a multi-modal sensor array to collect data on the temperature in each area of the annealing furnace, the speed, thickness, and microstructure evolution of the silicon steel strip at ultra-high frequencies. Preliminarily process the data at the sensor nodes through edge computing, and use fiber Bragg grating temperature sensing technology and self-calibrating measuring instruments to ensure high-precision and real-time data acquisition. Adopt multi-source data cross-validation to improve data reliability.

[0058] Deep spatio-temporal modeling prediction step: Input the collected multi-source data into a joint prediction model based on LSTM and GNN, and use an improved loss function for training and prediction. Utilize transfer learning to accelerate model convergence, and combine model compression technology to improve online operation efficiency, and accurately predict the temperature change during the annealing process.

[0059] Intelligent collaborative temperature control step: Adopt a composite control strategy of reinforcement learning and MPC, combined with fuzzy adaptive PID auxiliary control. According to the real-time temperature state and prediction results, use an improved power adjustment formula to adjust the power of the heating element in real time to achieve precise and efficient temperature control.

[0060] Panoramic real-time monitoring and feedback step: Through a distributed fiber optic sensing network, thermal imaging cameras, and multi-sensor data fusion, panoramically and real-time monitor the temperature of the annealing furnace and the quality parameters of the silicon steel. Once an abnormality is detected, it is quickly fed back to the temperature control step, and at the same time, the data is encrypted and stored in a distributed ledger database.

[0061] Multi-dimensional intelligent warning step: Rely on a multi-variable Bayesian network to build an accurate warning model, incorporating rich production data and failure cases. Use a carefully improved warning formula to comprehensively consider multiple factors to calculate warning indicators. Once the indicator exceeds the established threshold, the system immediately triggers an audible and visual alarm, quickly pushes warning information through the Internet of Things, and automatically generates a detailed fault diagnosis report.

[0062] Deep data insight and optimization step: Use a deep autoencoder and a generative adversarial network (GAN) to deeply mine production data. At the same time, combine association rule mining and clustering analysis techniques to accurately locate the key factors affecting production and sort out hidden rules. Based on the mining results, adaptively optimize the model structure and adjust the control strategy to improve the overall efficiency.

[0063] In the present invention, it also includes an adaptive dynamic parameter adjustment step. During the entire production process, the system closely monitors the real-time production conditions, and has data on equipment operating speeds, material input amounts, etc. at hand, and simultaneously collects product quality feedback. Based on this, use a genetic algorithm to simulate the biological evolution logic to accurately dynamically tune the model and control parameters, and then use an adaptive adjustment formula to optimize the parameters considering multiple factors to ensure that the system performance always maintains the optimal state.

[0064] In the present invention, there is also a remote holographic monitoring and operation step. By virtue of the ultra-high speed and low latency characteristics of the 5G network and advanced holographic projection technology, this step can accurately present a three-dimensional holographic image of the scene at the remote end, realizing remote holographic monitoring and operation as if on the spot. The strict permission management mechanism within the system accurately distributes operation permissions according to personnel responsibilities, task requirements, etc., preventing unauthorized actions. At the same time, the detailed operation log recording function is enabled, completely retaining information such as the time, personnel, instructions, and results of each remote operation, providing a solid basis for subsequent review and responsibility tracing, comprehensively ensuring the safety and reliability of remote operations, and effectively enhancing the intelligence and practicality of the system.

[0065] The above is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, making equivalent substitutions or changes, should be covered within the protection scope of the present invention.

Claims

1. An intelligent temperature control system for the continuous annealing process of non-oriented silicon steel, characterized in that, It includes the following modules: Multi-source perception fusion data acquisition module: Adopt a multi-modal sensor array including temperature, speed, thickness measurement and silicon steel microstructure monitoring. The temperature sensors are distributed by fiber Bragg gratings, the speed sensors and thickness gauges are self-calibrated, and the original data is preliminarily processed using edge computing technology; Deep spatio-temporal modeling module: Build a prediction model based on the long short-term memory network (LSTM) and graph neural network (GNN). The LSTM captures the dynamic changes of continuous annealing of silicon steel, and the GNN learns the heat transfer correlation in the spatial dimension; Intelligent collaborative temperature control module: Combining reinforcement learning with model predictive control (MPC), comprehensively considering control objectives and system constraints, and improving the formula Adjust the power, where α, β, and γ are dynamically adjusted coefficients, is the predicted temperature at the future k-th moment, and N is the prediction horizon; Panoramic real-time monitoring and feedback module: Deploy a distributed fiber optic sensing network and thermal imaging cameras to monitor the temperature of the annealing furnace and the silicon steel strip, and build a temperature monitoring system; Evaluate the quality through a multi-sensor fusion algorithm. When an anomaly occurs, the deviation is transmitted to the temperature control module for closed-loop control through a feedback mechanism, and the monitoring data is encrypted and stored in a distributed ledger database; Multi-dimensional intelligent early warning module: Establish an abnormal temperature early warning model of a multi-variable Bayesian network, construct a probabilistic causal network, and use the improved formula to calculate the early warning index, where ρ i is the weight of different abnormal events, P(A i |E) is the posterior probability of the occurrence of abnormal event A i under the current evidence E, and l is the number of abnormal events; When W exceeds the threshold, start the audible and visual alarm, push information to the mobile terminal through the Internet of Things, and automatically generate a fault diagnosis report; Deep data insight and optimization module: Use the technology combining deep autoencoders and generative adversarial networks to mine the monitoring and historical data, and use association rule mining and clustering analysis to find the reasons affecting the annealing temperature and silicon steel quality, and adaptively optimize according to the mining results.

2. The intelligent temperature control system for the non-oriented silicon steel continuous annealing process according to claim 1, characterized in that, It also includes an adaptive dynamic parameter adjustment module, which, based on production conditions and quality feedback, uses a genetic algorithm to adjust the LSTM and GNN network parameters in the deep spatio-temporal modeling module, as well as the reinforcement learning reward function and MPC prediction horizon parameters in the intelligent collaborative temperature control module, and applies the adaptive adjustment formula to optimize the model and control parameters, where θ old is the old parameter, δ is the adaptive learning rate, and is the gradient of the loss function with respect to the parameter.

3. The intelligent temperature control system for the non-oriented silicon steel continuous annealing process according to claim 1, characterized in that, It also includes a remote holographic monitoring and operation module. This module uses 5G communication technology and holographic projection technology to realize remote monitoring and operation of the annealing furnace temperature and production process. The operator can view the actual situation inside the annealing furnace in the form of a holographic image through a remote terminal device, support remote operation of heating element power adjustment and equipment start / stop functions, and have a hierarchical management of operation permissions and an operation log recording function.

4. The intelligent temperature control system for the non-oriented silicon steel continuous annealing process according to claim 1, characterized in that, The sensors in the multi-source perception fusion data acquisition module have self-diagnosis and self-repair functions. Through the built-in fault detection algorithm, the working state of the sensors is monitored. When a fault is found, the backup sensors are activated, and the MEMS technology and intelligent materials are used to automatically repair the faulty sensors. The multi-source data cross-validation technology is adopted to improve the data reliability.

5. The intelligent temperature control system for the non-oriented silicon steel continuous annealing process according to claim 1, characterized in that, The input of the deep spatio-temporal modeling module covers the chemical composition of silicon steel, strip speed, thickness and real-time temperature, and microstructure evolution data. Through training on historical and real-time production data, an improved loss function is used to optimize the model parameters, where is the predicted temperature, is the true temperature, ω i is the weight of different data points, λ is the regularization coefficient, and θ j are the model parameters; The transfer learning technique is adopted to pre-train the model using data from other similar metal heat treatment processes, and then fine-tune the continuous annealing data of non-oriented silicon steel; The model compression technique is used to reduce the model storage and calculation overhead.

6. The intelligent temperature control system for the non-oriented silicon steel continuous annealing process according to claim 1, wherein The intelligent collaborative temperature control module introduces fuzzy adaptive PID control as an auxiliary control strategy. When uncertainties occur in the system, it automatically switches to fuzzy adaptive PID control, and dynamically adjusts the PID controller parameters according to the temperature deviation and the rate of change of the deviation using fuzzy rules, and works in collaboration with reinforcement learning and MPC composite control.

7. The intelligent temperature control system for the non-oriented silicon steel continuous annealing process according to claim 1, characterized in that, The real-time monitoring and feedback module uses digital twin technology to build a virtual model of the annealing furnace and the silicon steel production process. By synchronizing with the actual production data, it simulates the annealing process, predicts potential problems, and provides a virtual verification environment for optimizing the control strategy.

8. A method for an intelligent temperature control system applying the non-oriented silicon steel continuous annealing process according to any one of claims 1-7, characterized in that, It includes the following steps: Multi-source perception data acquisition step: Use a multi-modal sensor array to collect the temperature of each area in the annealing furnace, the speed, thickness and microstructure evolution data of the silicon steel strip. The data is preliminarily processed at the sensor nodes through edge computing, and is collected in real time using fiber Bragg grating temperature sensing technology and self-calibrated measuring instruments, and the multi-source data cross-validation is adopted to improve the data reliability; Deep spatio-temporal modeling prediction steps: Input the collected multi-source data into the joint prediction model based on LSTM and GNN, train it using the improved loss function, utilize transfer learning to accelerate model convergence, and combine model compression technology to improve the online operation efficiency; Intelligent collaborative temperature control steps: Adopt the composite control strategy of reinforcement learning and MPC, combine fuzzy adaptive PID auxiliary control, and adjust the power of heating elements using the improved power adjustment formula according to the real-time temperature state and prediction results; Panoramic real-time monitoring and feedback steps: Through the distributed fiber optic sensing network, thermal imaging camera and multi-sensor data fusion, panoramically and real-time monitor the temperature of the annealing furnace and the quality parameters of silicon steel. Once an anomaly is found, it is fed back to the temperature control steps. At the same time, the data is encrypted and stored in the distributed ledger database; Multi-dimensional intelligent warning steps: Based on the multi-variable Bayesian network warning model, calculate the warning index using the improved warning formula. When the index exceeds the threshold, immediately trigger an alarm and push warning information, and at the same time generate a fault diagnosis report; Deep data insight and optimization steps: Use deep autoencoders and GAN to deeply mine the data, combine association rule mining and clustering analysis to discover key factors and patterns, and adaptively optimize the model and control strategy.

9. The method of the intelligent temperature control system for the non-oriented silicon steel continuous annealing process according to claim 8, characterized in that, It also includes an adaptive dynamic parameter adjustment step, which dynamically adjusts the model and control parameters using the genetic algorithm based on the real-time production conditions and quality feedback, and optimizes the parameters using the adaptive adjustment formula.

10. The method of the intelligent temperature control system for the non-oriented silicon steel continuous annealing process according to claim 8, characterized in that It also includes a remote holographic monitoring and operation step, which realizes remote holographic monitoring and operation with the help of 5G and holographic projection technology, and has functions of permission management and operation log recording.

Images