Online prediction and diagnosis method and system for process data of steel

Abstract

The invention relates to the technical field of data prediction, in particular to an online prediction and diagnosis method and system for process data of steel. The method comprises the following steps: deploying a distributed sensor network at a key station of a steel rolling production line, collecting rolling temperature gradient distribution, plate thickness fluctuation and rolling stress tensor, and constructing a multi-dimensional process characteristic data set; carrying out microcosmic physical constraint through the multi-dimensional process characteristic data set to generate multi-dimensional microcosmic physical constraint data; carrying out physical-data hybrid decoupling on the multi-dimensional microscopic physical constraint data, and carrying out digital twinborn model construction to obtain a steel prediction digital twinborn model; therefore, by comprehensively applying the distributed sensing technology, dynamic modeling and real-time data analysis, the limitation of process deviation and defect prediction in traditional steel production is solved, and the automation level and product consistency in the steel production process are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data prediction, and particularly to an online prediction and diagnosis method and system for process data of steel. Background Art

[0002] The prior art lacks the systematic real-time acquisition and dynamic analysis capabilities for multi-dimensional process data during the production process. In the traditional steel production process, although physical quantities such as temperature, pressure, and thickness are monitored through sensors, these data are often single and cannot be real-time fed back to the operators during the production process, resulting in the lag of production adjustment. In addition, the acquisition and analysis of process data are often based on preset rules and empirical models, lacking the deep learning and adaptation capabilities for complex and dynamically changing production environments and diversified factors. Secondly, the existing steel process data analysis methods have limitations in dealing with large-scale and multi-source data. With the development of industrial Internet of Things and intelligent sensor technologies, there are various types of process data generated during the steel production process, and the data volume is huge. Traditional analysis methods often rely on manually set models, and it is difficult to effectively integrate real-time data from different sources for comprehensive analysis, resulting in low prediction accuracy for process deviation degree, defect risk, and product quality, and unable to respond to abnormal changes in the production process in real time. Although existing diagnostic systems can identify some process anomalies, they lack the deep integration and comprehensive analysis of multi-dimensional data and cannot achieve full-process dynamic monitoring and early warning. In addition, the prediction and diagnosis technologies for steel production processes also have deficiencies in terms of adaptability and flexibility. Existing systems usually detect process deviations based on static rules and pre-set tolerance standards, and it is difficult to make adaptive adjustments according to the changes in real-time data. In a changing production environment, fixed rules cannot cope with complex process changes and external interferences. Summary of the Invention

[0003] Based on this, it is necessary to provide an online prediction and diagnosis method and system for process data of steel to solve at least one of the above technical problems.

[0004] To achieve the above object, an online prediction and diagnosis method for process data of steel, the method includes the following steps:

[0005] Step S1: Deploy a distributed sensing network at key workstations on the steel rolling production line, collect the rolling temperature gradient distribution, sheet thickness fluctuations, and roll pressure stress tensors, and construct a multi-dimensional process feature dataset;

[0006] Step S2: Perform microscopic physical constraints through the multi-dimensional process feature dataset to generate multi-dimensional microscopic physical constraint data; perform physical-data hybrid decoupling on the multi-dimensional microscopic physical constraint data, and construct a digital twin model to obtain a steel prediction digital twin model;

[0007] Step S3: Analyze the process parameter deviation degree according to the steel prediction digital twin model to generate steel process deviation degree data; perform dynamic correction of the tolerance boundary on the steel process deviation degree data to obtain the dynamic correction result of the steel; formulate a defect risk level based on the dynamic correction result of the steel, and perform process diagnosis on the steel through the defect risk level to obtain the steel process diagnosis result;

[0008] Step S4: Push the steel process diagnosis result to the cloud through the industrial Internet of Things platform for the closed-loop action of the diagnosis report to obtain the online prediction diagnosis report of the steel process.

[0009] The beneficial effects of the present invention are as follows. By deploying a distributed sensing network, various key process data during the rolling process (such as temperature gradient, sheet thickness fluctuation, and roll pressure stress tensor) are collected in real time, and a multi-dimensional feature data set is constructed, providing accurate data support for process optimization and anomaly detection during the steel production process. Through the coupling of the dynamic heat conduction equation and the material phase transformation kinetics model, the austenite conversion rate and grain growth trend can be accurately calculated. Combining the elastic-plastic constitutive equation with the real-time stress-strain data, the rolling force prediction value is iteratively updated. This process, by integrating the dislocation density model and the recrystallization kinetics model, can not only predict the critical conditions of dynamic / static recrystallization but also provide key data support for the construction of the digital twin model of steel, greatly enhancing the prediction ability of the steel production process. On this basis, the dynamic time warping algorithm (DTW) is used to analyze the process deviation degree of the steel prediction digital twin model, enabling the fluctuation of process parameters during the production process to be monitored in real time and compared with the standard process trajectory. The dynamic correction of the tolerance boundary ensures that the process deviation degree is always kept within an acceptable range, and then an accurate defect risk level and process diagnosis report are generated. These diagnosis results are pushed to the central control system through the industrial Internet of Things platform to trigger the closed-loop action of the diagnosis report in real time, ensuring that the defect risks during the production process are timely feedback and processed, thereby guaranteeing the production quality and process stability. This system forms a comprehensive, accurate, and real-time responsive steel process diagnosis and optimization platform by integrating a distributed sensing network, physical model coupling, digital twin technology, and dynamic time warping algorithm. It effectively solves the problems of lagging process anomaly detection, insufficient data analysis accuracy, and lack of real-time feedback mechanism in the traditional production process, improving the process control ability and product quality stability of steel production. Therefore, the present invention solves the limitations of process deviation and defect prediction in traditional steel production by comprehensively applying distributed sensing technology, dynamic modeling, and real-time data analysis, improving the automation level and product consistency of the steel production process.

[0010] Preferably, the construction of the multi-dimensional process feature data set includes the following steps:

[0011] The multi-spectral infrared thermal imaging array deployed through the distributed sensing network monitors the 0-15 mm gradient under the surface of the rolled piece according to the spectral range of 3-5 μm and the sampling frequency of 200 Hz, and obtains the rolling temperature gradient distribution;

[0012] According to the laser Doppler thickness measurement system, the initial inspection is carried out 10 m at the exit of the finishing mill, and the final inspection is carried out 5 m in front of the coiler. Scanning is carried out according to the scanning density of 256 points / section and the spacing of 8 mm to obtain the thickness fluctuation of the sheet;

[0013] The piezoelectric sensors deployed by the six-dimensional force sensing roll system collect the main rolling force component and the transverse stress component according to the high speed of 2 MHz and the wavelength demodulation frequency of 10 kHz, and obtain the roll pressure stress tensor;

[0014] The rolling temperature gradient distribution, the sheet thickness fluctuation and the roll pressure stress tensor are summarized and integrated into a data set to obtain a multi-dimensional process feature data set.

[0015] In the present invention, the multi-spectral infrared thermal imaging array deployed through the distributed sensing network uses a spectral range of 3-5 μm and a sampling frequency of 200 Hz to perform real-time monitoring on the 0-15 mm gradient under the surface of the rolled piece, and can accurately capture the change of the temperature gradient during the rolling process. This high-frequency data acquisition not only provides the thermal state information during the steel rolling process, but also provides high-quality input data for subsequent temperature field modeling and heat conduction analysis. The temperature gradient distribution is an important parameter for thermodynamic processes such as phase transformation, grain growth, and recrystallization of steel, and can help accurately calculate the austenite conversion rate and grain size, so as to optimize the quality and performance of steel. In addition, the laser Doppler thickness measurement system scans at two key positions, namely, at the exit of the finishing mill and in front of the coiler, according to the scanning density of 256 points / section and the spacing of 8 mm, to monitor the thickness fluctuation of the steel plate in real time. Through this system, accurate sheet thickness change data can be obtained, which provides a key basis for subsequent process optimization. The real-time feedback of the sheet thickness can directly affect the control of the rolling force, the roll pressure distribution, and the dimensional accuracy of the final steel product, and further optimize the material flow and pressing effect during the production process to ensure the consistency of the product quality. Furthermore, the piezoelectric sensors deployed by the six-dimensional force sensing roll system, combined with the high speed of 2 MHz and the wavelength demodulation frequency of 10 kHz, can collect the main rolling force component and the transverse stress component with high precision, and further accurately measure the roll pressure stress tensor. This data not only reveals the stress state during the rolling process, but also provides important parameter support for subsequent stress-strain analysis, dislocation density calculation, and recrystallization kinetics modeling. The real-time monitoring of the roll pressure stress tensor can help optimize the roll system pressure distribution and control the deviation of process parameters, and ensure the stability of the rolling process and the uniformity of the product.

[0016] Preferably, step S2 includes the following steps:

[0017] Step S21: plotting an austenite conversion rate curve for the rolling temperature gradient distribution to obtain an austenite conversion rate curve for the steel; and calculating a grain growth distribution diagram for the rolling temperature gradient distribution to obtain a grain size distribution cloud diagram;

[0018] Step S22: using the elastic-plastic constitutive equation to predict the rolling prediction value of the plate thickness fluctuation to obtain the plate rolling force prediction value;

[0019] Step S23: judging the critical recrystallization condition of the roller compression stress tensor to obtain critical recrystallization data of the steel;

[0020] Step S24: Perform micro-physical customized convolution based on the steel austenite conversion rate curve, grain size distribution cloud map, plate rolling force prediction value and steel recrystallization critical data to generate multi-dimensional micro-physical constraint data; construct a steel prediction digital twin model based on the multi-dimensional micro-physical constraint data to obtain a steel prediction digital twin model.

[0021] The present invention establishes a temperature influence model on the phase transformation process of steel based on the rolling temperature gradient distribution data by calculating the austenite conversion rate curve and the grain growth distribution diagram, and further provides the microstructure evolution information of steel at different rolling temperatures. These data provide key information for subsequent physical modeling and process optimization, and can analyze the transformation trend of steel under precise temperature and strain conditions, ensuring the meticulousness and accuracy of the prediction. Secondly, step S22 predicts the thickness fluctuation of the plate through the elastic-plastic constitutive equation, so as to accurately calculate the rolling force of the plate. This can not only help to grasp the deformation process of the plate, but also provide real-time feedback of the rolling force, ensuring effective control of the stress distribution and physical changes in the production process. The critical condition determination of the recrystallization of the roller stress tensor in step S23 provides key diagnostic data for the recrystallization behavior of steel in the production process. Through this step, the microstructure changes of the steel can be monitored and adjusted to avoid the situation of excessive or uneven grains, and optimize the quality and performance of the steel. Finally, step S24 forms a digital twin model of steel by performing micro-physical customized convolution on the austenite conversion rate, grain size distribution, rolling force prediction value and recrystallization critical data of the steel, and combining the multi-dimensional data to build the model. This process, through high integration at the data level, establishes a multi-dimensional digital platform that can reflect microscopic changes and provide real-time feedback on macroscopic process status, thereby achieving accurate prediction and optimized control of the steel production process.

[0022] Preferably, step S21 includes the following:

[0023] Step S211: Establish a non-uniform temperature gradient tensor for the rolling temperature gradient distribution, and perform unstructured grid discretization to obtain the three-dimensional temperature field of the steel; correlate the three-dimensional temperature field of the steel with the non-uniform temperature gradient tensor, with the concentration range being 100 - 200 kJ / mol, to obtain the phase transformation activation energy of the steel;

[0024] Step S212: Calculate the austenite conversion rate based on the phase transformation activation energy of the steel to obtain the austenite conversion rate data of the steel; perform curve transformation on the austenite conversion rate data of the steel to generate the austenite conversion rate curve of the steel;

[0025] Step S213: Calculate the grain boundary migration based on the non-uniform temperature gradient tensor, and analyze the grain growth trend between 5 μm and 50 μm to obtain the grain size distribution nephogram.

[0026] The present invention successfully constructs a three-dimensional temperature field of steel by transforming the rolling temperature gradient distribution into a non-uniform temperature gradient tensor and discretizing it using unstructured grids, providing key data for accurately describing and simulating the temperature distribution of steel during the rolling process. This high-resolution temperature field modeling ability can not only reflect the changes in temperature gradients at different rolling stages in real time but also reveal the influence of temperature on various physical properties and phase transformation processes of steel, laying a foundation for subsequent process optimization and performance prediction. By correlating the alloy element concentrations (concentration range: 100 - 200 kJ / mol) between the three-dimensional temperature field of steel and the non-uniform temperature gradient tensor, the phase transformation activation energy of steel is further obtained. This process can provide accurate thermodynamic data support for the phase transformation processes of different steel grades. The phase transformation activation energy is one of the core physical quantities during the rolling process of steel, and its value directly affects the phase transformation rate, grain growth, and austenite transformation process of steel, thus having an important impact on the quality and performance of the final product. Using this data, the austenite conversion rate of steel can be deduced, providing guidance for temperature control and phase transformation management in actual production. Based on the austenite conversion rate data, the austenite conversion rate curve of steel is generated through curve transformation, providing an intuitive analysis tool for the heat treatment process, material composition optimization, and performance regulation of steel. These curves can help the production line accurately monitor and adjust the rolling process parameters to ensure the uniformity and consistency of the product under different rolling conditions and reduce quality fluctuations caused by temperature fluctuations and composition changes. In addition, the grain boundary migration is calculated according to the non-uniform temperature gradient tensor, and the grain growth trend is analyzed (grain size range: 5 μm to 50 μm), further obtaining the cloud map of the grain size distribution of steel. This analysis not only reveals the microstructural changes of steel during the rolling process but also provides an important basis for controlling grain refinement and improving the mechanical properties of materials. The grain size directly affects the strength, toughness, and other mechanical properties of steel. Therefore, accurate grain size analysis is of great significance for quality control during the production process and the improvement of the performance of the final product.

[0027] Preferably, step S22 includes the following:

[0028] Step S221: Perform an equivalent plastic strain increment dual-field analysis on the sheet thickness fluctuation to obtain the sheet thickness-strain field information;

[0029] Step S222: Construct a thickness fluctuation constraint elastoplastic equation for the sheet thickness-strain field information and deduce the mixed-driving constitutive parameters to obtain the sheet mixed-driving constitutive parameter data;

[0030] Step S223: Predict the rolling force for the sheet mixed-driving constitutive parameter data to obtain the predicted value of the sheet rolling force.

[0031] Through the equivalent plastic strain increment dual-field analysis of the sheet thickness fluctuation, the sheet thickness-strain field is obtained, providing key mechanical behavior data for subsequent process parameter optimization. The equivalent plastic strain increment dual-field analysis can not only reveal the strain distribution and thickness fluctuation of the sheet during rolling, but also help identify material inhomogeneities caused by factors such as temperature, pressure, and speed, ensuring a deeper understanding of the plastic deformation behavior of the sheet during complex rolling processes. Through this analysis method, the stress-strain state of the sheet during rolling can be captured more accurately, and thus a more precise model basis can be provided for optimizing the production process. On this basis, an elastoplastic equation with thickness fluctuation constraints is constructed, and the constitutive parameters of hybrid drive are deduced. This step further improves the prediction accuracy of the sheet deformation process by considering the influence of sheet thickness fluctuation on plastic deformation. The elastoplastic equation with thickness fluctuation constraints can comprehensively consider the local stress concentration phenomenon caused by the uneven material thickness during rolling and introduce the complex stress state existing in the actual working conditions. This equation can provide an effective physical model for the accurate prediction of the subsequent rolling force, ensuring that the obtained results have high practical application value. Subsequently, by using the Newton iteration method with thickness feedback to predict the rolling force for the constitutive parameter data of the sheet hybrid drive, the rolling force parameters can be adjusted in real time, thereby improving the accuracy and stability during the rolling process. The application of the Newton iteration method in this step can quickly and accurately adjust the constitutive parameters according to the thickness and strain information, and then optimize the prediction of the rolling force, making the pressure distribution during the rolling process more uniform and avoiding sheet defects and quality problems caused by uneven stress. In addition, this prediction process can not only ensure the uniformity of the sheet thickness, but also reduce the energy consumption and material waste during the rolling process and improve the production efficiency.

[0032] Preferably, step S23 includes the following:

[0033] Step S231: Extract the characteristics of the hydrostatic stress tensor according to the rolling stress tensor, and calculate the equivalent stress to obtain the rolling equivalent stress data;

[0034] Step S232: Evolve the dislocation density according to the rolling equivalent stress data to generate rolling stress-driven evolution data; perform intelligent condition determination on the rolling stress-driven evolution data to obtain the recrystallization critical data.

[0035] Through the extraction of the hydrostatic stress tensor characteristics from the rolling stress tensor, the equivalent stress is then calculated to obtain the rolling equivalent stress data, which plays a crucial role in the stress distribution analysis during the steel rolling process. The extraction of hydrostatic stress tensor characteristics can transform the complex multi-dimensional stress field into easily understandable and calculable equivalent stress data, which provides a scientific basis for further analyzing the plastic deformation, material flow, and quality control of materials during the rolling process. As a key parameter during the material deformation process, the equivalent stress directly affects the mechanical properties, microstructure, and the quality of the final product of the steel. Through this data-driven method, the stress state during the rolling process can be accurately grasped, providing a basis for further process optimization. Based on the rolling equivalent stress data, the calculation of the dislocation density evolution is carried out to generate the rolling stress-driven evolution data. This data can not only reflect the microstructure changes of the steel during the rolling process but also reveal the movement and evolution process of dislocations caused by stress. The dislocation density is a fundamental indicator of material deformation and strength. Understanding its evolution during the rolling process helps to deeply understand the plastic rheological behavior of the material. By dynamically calculating the change in the dislocation density, the internal micro-defects and the changes in the lattice structure of the steel can be monitored in real time, which is crucial for predicting the mechanical properties, strength, and toughness of the material. Through the intelligent condition determination of recrystallization triggering for the rolling stress-driven evolution data, the dynamic / static recrystallization critical data is obtained, further providing accurate reference for the heat treatment and grain refinement process of the steel. This determination process is realized through intelligent algorithms, which can evaluate the recrystallization state of the steel in real time during the complex rolling process and dynamically adjust the process parameters according to the actual situation to ensure the best performance of the steel during the hot working process. The dynamic / static recrystallization critical data provides strong support for subsequent process optimization, heat treatment plan design, and product quality control.

[0036] Preferably, the intelligent condition determination of recrystallization triggering includes the following:

[0037] When the real-time dislocation density ρ t is greater than or equal to ρ crit the intelligent condition determination is triggered; where ρ crit is calculated as follows:

[0038]

[0039] In the formula, ρ crit is the critical dislocation density, σ eq is the equivalent stress, α is the material constant, G is the shear modulus, b is the Burgers vector modulus, is the plastic strain rate, is the critical strain rate.

[0040] By establishing a dynamic criterion based on the critical dislocation density, the present invention significantly improves the spatio-temporal resolution and process adaptability of recrystallization process prediction. Its core innovation lies in the non-linear coupling of the equivalent stress field and the strain rate field. By introducing the hyperbolic tangent function to construct the strain rate sensitivity factor, it breaks through the limitations of the traditional static critical strain criterion in the dynamic rolling scenario, enabling the accurate capture of the phase transformation inflection point of dislocation density evolution under the non-steady processing conditions of severe deformation. By real-time fusing multi-source sensor data (such as the equivalent stress σ eq analyzed from the roll pressure stress tensor, and the plastic strain rate inverted by laser thickness measurement), combined with the intrinsic parameters (shear modulus G, Burgers vector b) in the material database, this model realizes the cross-scale correlation between the microstructure evolution and the macroscopic mechanical state, providing a quantitative determination benchmark for the triggering time of dynamic recrystallization.

[0041] Preferably, the obtaining of the steel prediction digital twin model includes the following steps:

[0042] Obtain a preset digital twin model;

[0043] Extract the austenite conversion rate of the steel according to the austenite conversion rate curve of the steel, and perform the treatment of the transformation of austenite in the steel into other phases to generate the austenite phase transformation parameters of the steel, and finally inject them into the sub-module of the preset digital twin model for the prediction of the phase transformation process to obtain the phase transformation and grain evolution module;

[0044] Analyze the microstructure uniformity and local variation of the grain size distribution cloud map to generate grain micro-feature data; perform the real distribution calculation on the grain micro-feature data, and finally perform the reconstruction of the micro-grain structure and generate the grain module to obtain the grain structure and distribution module;

[0045] Perform module fusion according to the phase transformation and grain evolution module and the grain structure and distribution module to generate a preliminary digital twin model; organically combine the predicted value of the sheet rolling force and the recrystallization critical data of the preliminary digital twin model, and perform model generation to obtain a preliminary fusion digital twin model;

[0046] Perform cross-module information integration on the preliminary fusion digital twin model and adjust the module parameters to obtain the steel prediction digital twin model.

[0047] Through steps such as the presetting of the digital twin model, the extraction of the austenite conversion rate, the analysis of the grain size distribution cloud map, and module fusion, the present invention effectively improves the process control accuracy and prediction ability during the steel rolling process. First, by obtaining the preset digital twin model, a digital and dynamic platform for the steel rolling process is established, providing a reliable framework for subsequent process parameter prediction and optimization. On this basis, the austenite conversion rate is extracted according to the steel austenite conversion rate curve, and the process of austenite phase transformation into other phases is processed, injecting relevant phase transformation parameters into the sub-modules of the digital twin model, thereby realizing the real-time prediction of the phase transformation process. This process enables the accurate prediction of the phase transformation behavior of steel during the rolling process, especially during the heat treatment stage, providing key data support for the quality control of the final product. At the same time, through the analysis of the microstructure uniformity and local variation of the grain size distribution cloud map, combined with the calculation of the true distribution of the grain microscopic characteristic data, the microscopic structure evolution of steel during the rolling process can be deeply explored, and the mechanical properties and microscopic tissue structure of steel can be further optimized through the reconstruction of the microscopic grain structure. During the process of fusing the phase transformation and grain evolution module and the grain structure and distribution module, not only the data sharing and synergy between different modules are realized, but also the heat treatment process of steel and the grain growth behavior are effectively integrated, thereby providing strong data support for the prediction of the rolling force of the sheet and the dynamic analysis of the recrystallization critical conditions. Through this multi-module fusion method, the changes in the macroscopic properties and microscopic structure of the material can be considered simultaneously in the digital twin model, thereby improving the accuracy and stability during the rolling process. Finally, through cross-module information integration of the preliminarily fused digital twin model and adjusting the module parameters according to the actual working conditions, a more accurate and reliable digital twin model for steel prediction is obtained. This model not only improves the accuracy of steel process control, but also can reflect in real time the influence of process parameter changes on the mechanical properties and microscopic tissue structure of steel, providing strong technical support for the quality control, process optimization, and production efficiency improvement of steel.

[0048] Preferably, step S3 includes the following steps:

[0049] Step S31: Analyze the process parameter deviation degree of the digital twin model for steel prediction by using the dynamic time warping algorithm to obtain the steel process deviation degree data;

[0050] Step S32: Perform data point matching and distance calculation according to the steel process deviation degree data to obtain the steel time series deviation degree map;

[0051] Step S33: Dynamically correct the tolerance boundary of the steel time series deviation degree map to obtain the steel dynamic correction result; formulate the defect risk level based on the steel dynamic correction result, and use the defect risk level to perform process diagnosis on the steel to obtain the steel process diagnosis result.

[0052] By adopting the dynamic time warping algorithm, the present invention analyzes the deviation degree of process parameters of the steel prediction digital twin model, and gradually performs data point matching, distance calculation and dynamic correction of tolerance boundaries, providing an accurate technical means for quality control and process optimization in the steel production process. First of all, the dynamic time warping algorithm can accurately identify the deviation situation in the process according to the process parameter data in the steel prediction digital twin model, and quantify the deviation degree data to ensure the timely monitoring and diagnosis of abnormal situations in the production process. The deviation degree data reflects various fluctuations and unstable factors in the process, and it can effectively reveal the process deviation sources and their potential impacts on the quality of the final product. Therefore, this data becomes an important basis for subsequent process adjustment and optimization. On this basis, by using the method of data point matching and distance calculation, the sequential changes of the steel process are further analyzed to generate a steel sequential deviation degree map. Through this map, the changing trends of process parameters at different time nodes can be clearly displayed, and the process deviations accumulated over time and their development trends can be revealed. This sequential deviation degree map not only provides visual support for real-time monitoring in the production process, but also provides historical data basis for future production decisions, enabling potential risks and problems in each link of the production process to be foreseen and intervened in advance. Finally, by dynamically correcting the steel sequential deviation degree map and the tolerance boundary, the fluctuation range of the production process can be effectively adjusted, thereby improving production stability and product consistency. Through the dynamic adjustment of the tolerance boundary, the risk of quality defects caused by process fluctuations can be minimized on the premise of ensuring production quality. In addition, the generated defect risk level and the steel process diagnosis result provide a clear risk assessment basis for decision-makers, helping them take targeted measures for correction and optimization in a timely manner. The data analysis, deviation correction and risk assessment links in this process provide a more intelligent and precise management solution for steel production.

[0053] In this specification, an online prediction and diagnosis system for process data of steel is provided, which is used to execute the above-mentioned online prediction and diagnosis method for process data of steel. The online prediction and diagnosis system for process data of steel includes:

[0054] A data acquisition and feature construction module deploys a distributed sensing network at key stations of the steel rolling production line to collect the rolling temperature gradient distribution, sheet thickness fluctuation and roll pressure stress tensor, and constructs a multi-dimensional process feature data set;

[0055] A physical model coupling and digital twin model generation module performs microscopic physical constraints through the multi-dimensional process feature data set to generate multi-dimensional microscopic physical constraint data; performs physical-data hybrid decoupling on the multi-dimensional microscopic physical constraint data, and constructs a digital twin model to obtain a steel prediction digital twin model;

[0056] The process deviation analysis and tolerance correction module analyzes the process parameter deviation degree according to the steel prediction digital twin model, generates steel process deviation degree data; dynamically corrects the tolerance boundary of the steel process deviation degree data to obtain the dynamic correction result of the steel; formulates the defect risk level based on the dynamic correction result of the steel, and conducts process diagnosis on the steel through the defect risk level to obtain the steel process diagnosis result;

[0057] The data push and diagnosis report generation module pushes the steel process diagnosis result to the cloud through the industrial Internet of Things platform for the closed-loop action of the diagnosis report, and obtains the online prediction diagnosis report of the steel process.

[0058] The beneficial effects of the present invention are as follows: by deploying a distributed sensing network to collect various key process data (such as temperature gradient, sheet thickness fluctuation, and roll pressure stress tensor) during the rolling process in real time, and constructing a multi-dimensional feature data set, it provides accurate data support for process optimization and anomaly detection in the steel production process. Through the coupling of the dynamic heat conduction equation and the material phase transformation kinetics model, the austenite conversion rate and grain growth trend can be accurately calculated. Combining the elastic-plastic constitutive equation with the real-time stress-strain data, the rolling force prediction value is iteratively updated. This process not only predicts the critical conditions of dynamic / static recrystallization by integrating the dislocation density model and the recrystallization kinetics model, but also provides key data support for the construction of the digital twin model of steel, greatly enhancing the prediction ability of the steel production process. On this basis, the dynamic time warping algorithm (DTW) is used to analyze the process deviation degree of the steel prediction digital twin model, enabling the fluctuation of process parameters during the production process to be monitored in real time and compared with the standard process trajectory. The dynamic correction of the tolerance boundary ensures that the process deviation degree is always kept within an acceptable range, and then generates accurate defect risk levels and process diagnosis reports. These diagnosis results are pushed to the central control system through the industrial Internet of Things platform to trigger the closed-loop action of the diagnosis report in real time, ensuring that the defect risks in the production process are timely feedback and processed, thus guaranteeing the production quality and process stability. By integrating a distributed sensing network, physical model coupling, digital twin technology, and the dynamic time warping algorithm, this system forms a comprehensive, accurate, and real-time responsive steel process diagnosis and optimization platform. It effectively solves the problems of lag in process anomaly detection, insufficient data analysis accuracy, and lack of real-time feedback mechanism in the traditional production process, and improves the process control ability and product quality stability of steel production. Therefore, the present invention solves the limitations of process deviation and defect prediction in traditional steel production by comprehensively applying distributed sensing technology, dynamic modeling, and real-time data analysis, and improves the automation level and product consistency of the steel production process. Description of the Drawings

[0059] Figure 1Schematic diagram of the step process of an online prediction and diagnosis method for process data of a steel material

[0060] Figure 2 is Figure 1 Schematic diagram of the detailed implementation steps of step S3 in

[0061] The realization, functional characteristics and advantages of the purpose of the present invention will be further described with reference to the embodiments and the accompanying drawings. Specific implementation manners

[0062] The technical method of the present invention for a patent will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of 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.

[0063] In addition, the accompanying drawings are only schematic diagrams of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings represent the same or similar parts, and thus the repeated description thereof will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor methods and / or microcontroller methods.

[0064] It should be understood that although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly, the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the associated listed items.

[0065] To achieve the above object, please refer to Figures 1 to 2 , an online prediction and diagnosis method for process data of a steel material, the method comprising the following steps:

[0066] Step S1: Deploy a distributed sensing network at key workstations on the steel rolling production line, collect the rolling temperature gradient distribution, plate thickness fluctuations and roll pressure stress tensors, and construct a multi-dimensional process feature dataset;

[0067] Step S2: Perform microphysical constraints through the multi-dimensional process feature dataset to generate multi-dimensional microphysical constraint data; perform physical-data hybrid decoupling on the multi-dimensional microphysical constraint data and construct a digital twin model to obtain a steel prediction digital twin model;

[0068] Step S3: Analyze the process parameter deviation degree according to the steel prediction digital twin model to generate steel process deviation degree data; perform tolerance boundary dynamic correction on the steel process deviation degree data to obtain a steel dynamic correction result; formulate a defect risk level based on the steel dynamic correction result and perform process diagnosis on the steel through the defect risk level to obtain a steel process diagnosis result;

[0069] Step S4: Push the steel process diagnosis result to the cloud through the industrial Internet of Things platform for closed-loop action of the diagnosis report to obtain an online prediction diagnosis report of the steel process.

[0070] The beneficial effects of the present invention are as follows. By deploying a distributed sensing network to collect various key process data (such as temperature gradient, sheet thickness fluctuation, and roll pressure stress tensor) during the rolling process in real time, and constructing a multi-dimensional feature data set, it provides accurate data support for process optimization and anomaly detection in the steel production process. Through the coupling of the dynamic heat conduction equation and the material phase transformation kinetics model, the austenite conversion rate and the grain growth trend can be accurately calculated. Combining the elastoplastic constitutive equation with the real-time stress-strain data, the rolling force prediction value is iteratively updated. This process, by integrating the dislocation density model and the recrystallization kinetics model, can not only predict the critical conditions of dynamic / static recrystallization, but also provides key data support for the construction of the digital twin model of steel, greatly enhancing the prediction ability of the steel production process. On this basis, the dynamic time warping algorithm (DTW) is used to analyze the process deviation degree of the steel prediction digital twin model, enabling the fluctuation of process parameters during the production process to be monitored in real time and compared with the standard process trajectory. The dynamic correction of the tolerance boundary ensures that the process deviation degree always remains within an acceptable range, and then generates accurate defect risk levels and process diagnosis reports. These diagnostic results are pushed to the central control system through the industrial Internet of Things platform, triggering the closed-loop action of the diagnostic report in real time to ensure that the defect risks in the production process are promptly feedback and processed, thereby ensuring the stability of production quality and process. This system forms a comprehensive, accurate, and real-time responsive steel process diagnosis and optimization platform by integrating a distributed sensing network, physical model coupling, digital twin technology, and the dynamic time warping algorithm. It effectively solves the problems of lag in process anomaly detection, insufficient data analysis accuracy, and lack of real-time feedback mechanism in the traditional production process, improving the process control ability and product quality stability of steel production. Therefore, the present invention solves the limitations of process deviation and defect prediction in traditional steel production by comprehensively applying distributed sensing technology, dynamic modeling, and real-time data analysis, improving the automation level and product consistency of the steel production process.

[0071] In an embodiment of the present invention, with reference to Figure 1 as shown, in this example, the online prediction and diagnosis method for the process data of the steel includes the following steps:

[0072] Step S1: Deploy a distributed sensing network at key stations on the steel rolling production line, collect the rolling temperature gradient distribution, sheet thickness fluctuation, and roll pressure stress tensor, and construct a multi-dimensional process feature data set;

[0073] In the embodiments of the present invention, a distributed sensing network is deployed at key workstations in the steel rolling production line, aiming to collect various key process parameters required during the rolling process in real time. Through the sensor array, data such as the rolling temperature gradient distribution, the fluctuation of the sheet thickness, and the roll pressure stress tensor are respectively monitored and recorded. These sensors select suitable working frequencies, measurement accuracies, and distribution positions according to different physical quantities and measurement requirements to ensure the comprehensiveness and accuracy of the data. The rolling temperature gradient distribution is mainly monitored in real time by devices such as infrared thermal imaging and thermocouple sensors, providing temperature change information at various positions of the steel during the rolling process and providing basic data for subsequent phase transformation analysis and temperature field simulation; the fluctuation of the sheet thickness is accurately monitored by technical means such as laser scanning and sensors to monitor the thickness changes on the surface and inside of the sheet, thereby reflecting the deformation generated during the rolling process and the actual thickness fluctuation of the sheet; the roll pressure stress tensor is measured in real time by strain sensors or mechanical sensors, recording the stress distribution of the steel during the rolling process, especially the stress changes in the main direction and the transverse direction, which is crucial for understanding the rolling force and the stress state of the sheet. Through the collaborative work of these sensor networks, the collected data forms a multi-dimensional process feature dataset, which provides rich basic data for subsequent process optimization, process prediction, and digital twin model construction.

[0074] Step S2: Perform microscopic physical constraints through the multi-dimensional process feature dataset to generate multi-dimensional microscopic physical constraint data; perform physical-data hybrid decoupling on the multi-dimensional microscopic physical constraint data and construct a digital twin model to obtain a steel prediction digital twin model;

[0075] In the embodiments of the present invention, based on the obtained multi-dimensional process feature dataset, the extraction of microscopic physical constraints is carried out. Specifically, various physical quantities such as temperature gradient, sheet thickness fluctuation, and rolling stress tensor can be used to establish a microscopic physical behavior model of steel during the rolling process. These data form a set of microscopic physical constraints describing the behavior of steel through the interaction of physical principles such as thermodynamics, mechanics, and materials science. By processing these process parameters and optimizing the analysis of physical constraints, a set of multi-dimensional microscopic physical constraint data is generated, covering aspects such as the thermal deformation, stress-strain state of steel, and its phase transformation behavior under different process conditions. Then, the system decouples these microscopic physical constraint data from the relevant physical models. Through physical-data hybrid decoupling technology, the system separates the relationship between physical constraints and experimental data obtained by sensors, and then effectively extracts independent and operable physical variables. This decoupling process usually transforms complex physical problems into more easily processed mathematical problems through algorithms such as principal component analysis (PCA) and Lagrange multiplier method, thus providing a simplified and accurate basis for subsequent modeling. Finally, based on these decoupled microscopic physical constraint data, the system constructs a digital twin model of steel. The construction of the digital twin model mainly relies on the close integration of physics and data, ensuring that the simulation process accurately reflects the actual rolling process.

[0076] Step S3: Analyze the process parameter deviation degree according to the steel prediction digital twin model to generate steel process deviation degree data; perform dynamic correction of the tolerance boundary on the steel process deviation degree data to obtain the steel dynamic correction result; formulate a defect risk level based on the steel dynamic correction result, and perform process diagnosis on the steel through the defect risk level to obtain the steel process diagnosis result;

[0077] In the embodiments of the present invention, by monitoring the multi-dimensional process characteristic data such as the thermal deformation, stress-strain state, and phase transformation behavior of steel in real time, the process parameters of steel during rolling are obtained and accurately predicted. However, in the actual production process, due to various external factors (such as equipment aging, operation errors, environmental changes, etc.), the process parameters often deviate. At this time, through the deviation analysis based on the steel prediction digital twin model, the system can dynamically calculate the deviation degree data of each process parameter, revealing which parameters have changed significantly and the degree of deviation from the expected range. Next, for the deviation degree of these process parameters, the system adjusts and optimizes the process data through the tolerance boundary dynamic correction algorithm to compensate for the deviation and reduce its negative impact on the production process. The correction of the tolerance boundary is based on specific process standards and real-time data analysis, dynamically adjusting the process range to obtain the dynamic correction result of the steel. These corrected data can provide a more accurate basis for subsequent defect prediction and process optimization. Based on the above dynamic correction results, the system further combines the defect risk assessment model to diagnose the process state of the steel and formulate the defect risk level. Specifically, the system classifies and evaluates the data of different deviation degrees and the corrected process parameters, and divides the process status of the steel into different risk levels, such as low risk, medium risk, and high risk levels. Finally, combined with the process diagnosis results of the steel, the system can propose corresponding process improvement measures or early warning strategies to avoid the occurrence of defects in production, thus providing a scientific basis for the quality control and optimization of the steel production process.

[0078] Step S4: Push the steel process diagnosis result to the cloud through the industrial Internet of Things platform for the closed-loop action of the diagnosis report to obtain the online prediction diagnosis report of the steel process.

[0079] In the embodiments of the present invention, the role of the industrial Internet of Things (IIoT) platform in this process is to build an efficient and low-latency communication architecture to achieve real-time collection, processing, and transmission of process data during steel production. By integrating various sensors and devices, such as temperature sensors, thickness sensors, stress sensors, etc., the platform can collect process parameter data during the steel rolling process in real time, preprocess and analyze this data, and generate diagnostic results. Since the industrial Internet of Things can connect different devices, sensors, and computing units, it provides an efficient data transmission channel, ensuring the seamless flow and real-time update of diagnostic data, thus ensuring that the central control system can obtain accurate process diagnostic information in a timely manner. Then, the data will go through necessary encryption and verification steps during transmission to ensure the security and integrity of the data and avoid loss or tampering during transmission. Through real-time monitoring and fault detection of the data, the platform can ensure that any process deviation during the steel rolling process can be detected and reported in a timely manner. At the same time, the IIoT platform will also integrate various algorithms and data processing tools to deeply analyze the collected diagnostic data and generate accurate diagnostic results, which include but are not limited to information such as process deviation and defect risk level. After transmitting these results to the central control system through the Internet of Things platform, the system will perform automated processing according to preset rules and logics and trigger relevant closed-loop actions, such as adjusting rolling parameters, scheduling production equipment, or optimizing the production process, so as to quickly respond to any problems in the process. Finally, the industrial Internet of Things platform realizes the generation of an online prediction diagnostic report for the steel process through accurate data analysis and an automated feedback mechanism. This report not only provides real-time process status assessment for operators but also can predict the future evolution trend of the process based on the combination of historical data and real-time data, thus providing a warning function.

[0080] Preferably, the construction of the multi-dimensional process feature dataset in step S1 includes the following:

[0081] The multi-spectral infrared thermal imaging array deployed through the distributed sensing network monitors the 0-15 mm gradient under the surface of the rolled piece in depth according to the spectral range of 3-5 μm and the sampling frequency of 200 Hz to obtain the rolling temperature gradient distribution;

[0082] According to the initial inspection of the laser Doppler thickness measurement system at 10 m from the exit of the finishing mill and the final inspection at 5 m before the coiler, scan according to the scanning density of 256 points / section and the spacing of 8 mm to obtain the thickness fluctuation of the sheet;

[0083] Summarize and integrate the rolling temperature gradient distribution, the thickness fluctuation of the sheet, and the roll pressure stress tensor into a dataset to obtain a multi-dimensional process feature dataset;

[0084] The rolling temperature gradient distribution, sheet thickness fluctuations, and roll pressure stress tensors are aggregated and integrated using a distributed sensing network to obtain a multi-dimensional process feature dataset.

[0085] In the embodiments of the present invention, a distributed sensing network, a multi-spectral infrared thermal imaging array, a laser Doppler thickness measurement system, and a six-dimensional force sensing roll system are used to comprehensively monitor and collect key process data during rolling, thereby constructing a multi-dimensional process feature dataset. The common goal of these technical means is to ensure the precise control and optimization of various parameters during the steel rolling process through accurate real-time data collection and analysis. First, the multi-spectral infrared thermal imaging array in the distributed sensing network monitors thermal radiation according to its spectral range (3 - 5μm), mainly used to capture the temperature gradient distribution on the surface of the rolled piece and at a depth of 0 - 15mm below the surface. By selecting this specific wavelength range, the multi-spectral infrared thermal imaging array can efficiently detect and distinguish different temperature regions, thereby providing detailed data support for subsequent thermal management and temperature control. In addition, the setting of a sampling frequency of 200Hz ensures the real-time and high-precision nature of this data, helping to accurately capture minute temperature fluctuations and changes during the rolling process, and thus providing a basis for optimizing the temperature control strategy. Next, the laser Doppler thickness measurement system measures the thickness fluctuations of the sheet through precise laser scanning technology. The system conducts a preliminary inspection 10 meters from the exit of the finishing mill and a final inspection 5 meters before the coiler, ensuring the coverage and reliability of the data. During the measurement process, the setting of a scanning density of 256 points / section ensures that each section can be sampled for thickness with high precision, and the scanning interval of 8mm enables this data to not only have a high spatial resolution but also provide an all-round monitoring of the thickness fluctuations of the sheet. This is crucial for controlling the stability of the rolling thickness and can effectively detect production deviations. In addition, the six-dimensional force sensing roll system uses piezoelectric sensors deployed on the rolling mill roll system to collect the main components of the rolling force and the lateral stress components in real time. The working frequency of the sensor is 2MHz, and the demodulation frequency is 10kHz, ensuring a high-frequency and high-precision data collection ability, capable of capturing minute changes in the stress state during the high-speed rolling process. The collection of the roll pressure stress tensor provides basic data for subsequent stress analysis and optimization, can accurately reflect the stress state of the material during the rolling process, and provides data support for optimizing the rolling force and roll system configuration.

[0086] Preferably, step S2 includes the following steps:

[0087] Step S21: Plot the austenite conversion rate curve for the rolling temperature gradient distribution to obtain the steel austenite conversion rate curve; deduce the grain growth distribution map for the rolling temperature gradient distribution to obtain the grain size distribution cloud map;

[0088] Step S22: Use the elastoplastic constitutive equation to predict the rolling prediction value of the sheet thickness fluctuation, and obtain the predicted value of the sheet rolling force;

[0089] Step S23: Judge the recrystallization critical condition of the roll pressure stress tensor to obtain the critical data of steel recrystallization;

[0090] Step S24: Perform microphysical customized convolution based on the austenite transformation rate curve of steel, the grain size distribution nephogram, the predicted value of the sheet rolling force, and the critical data of steel recrystallization to generate multi-dimensional microphysical constraint data; construct a digital twin model for steel prediction based on the multi-dimensional microphysical constraint data to obtain the digital twin model for steel prediction.

[0091] In the embodiment of the present invention, by analyzing the rolling temperature gradient distribution, the relationship between temperature and steel phase transformation is used to draw the austenite conversion rate curve, and the grain growth distribution map is deduced, so as to obtain the austenite conversion rate and the grain size distribution nephogram of steel during the rolling process. This process realizes the accurate prediction of the microstructure evolution of steel by combining the dynamic relationship between the temperature gradient and the steel phase transformation behavior. Step S22 uses the elastoplastic constitutive equation to model the sheet thickness fluctuation, so as to calculate the predicted value of the sheet rolling force, and then analyze the stress and deformation behavior of the sheet during the rolling process. This prediction can help optimize the rolling parameters and ensure the finished product quality. Step S23 further judges the recrystallization critical condition of the steel by analyzing the roll pressure stress tensor to generate the critical data of recrystallization. This step is to understand whether the steel reaches the critical state of recrystallization during the rolling process, and then make an effective prediction of the microstructure change of the steel. Finally, step S24 combines the austenite conversion rate curve, the grain size distribution nephogram, the predicted value of the rolling force, and the critical data of recrystallization to perform microphysical customized convolution to generate multi-dimensional microphysical constraint data.

[0092] Preferably, step S21 includes the following:

[0093] Step S211: Establish a non-uniform temperature gradient tensor for the rolling temperature gradient distribution, and perform unstructured grid discretization to obtain the three-dimensional temperature field of the steel; correlate the three-dimensional temperature field of the steel with the non-uniform temperature gradient tensor, and the concentration range is 100 - 200 kJ / mol to obtain the phase transformation activation energy of the steel;

[0094] Step S212: Deduce the austenite conversion rate according to the phase transformation activation energy of the steel to obtain the austenite conversion rate data of the steel; perform curve transformation on the austenite conversion rate data of the steel to generate the austenite conversion rate curve of the steel;

[0095] Step S213: Calculate the grain boundary migration according to the non-uniform temperature gradient tensor, and analyze the grain growth trend between 5 μm and 50 μm to obtain the grain size distribution nephogram.

[0096] In the embodiments of the present invention, by constructing and analyzing a mathematical model of the thermodynamic and mechanical changes suffered by steel during rolling, the phase transformation process, grain growth, and the evolution of its microstructure of the steel are accurately predicted. By establishing the rolling temperature gradient distribution and generating a non-uniform temperature gradient tensor, the technical means rely on thermodynamic equations and the unstructured grid discretization method to obtain the detailed distribution of the three-dimensional temperature field of the steel. This technical means effectively simulates the spatial variation of the temperature during the rolling process of the steel, provides detailed temperature field data, and further provides a basis for subsequent phase transformation and microstructure analysis. The discretization method of the unstructured grid can be flexibly applied in complex geometric bodies, thereby improving the accuracy and applicability of the model. The three-dimensional temperature field of the steel is combined with the non-uniform temperature gradient tensor to correlate the alloy element concentration, and the concentration range is 100 - 200 kJ / mol. Through this process, the phase transformation activation energy of the steel is calculated. This activation energy is one of the core parameters of the steel phase transformation and determines the phase transformation rate and type of the material at different temperatures. Through the correlation with the concentration data, the technical means can accurately calculate the phase transformation activation energy according to different alloy compositions, providing data support for the subsequent calculation of the austenite conversion rate. During the calculation of the austenite conversion rate, using the phase transformation activation energy and temperature information, the austenite conversion rate of the steel is accurately calculated through a thermodynamic model, generating austenite conversion rate data. These data are transformed into a curve to obtain the austenite conversion rate curve of the steel, further providing a quantitative description of the phase transformation process of the material. The austenite conversion rate curve of the steel helps to deeply understand the phase transformation behavior of the steel during rolling, especially how it affects the final properties of the material under different rolling conditions. Finally, based on the non-uniform temperature gradient tensor, grain boundary migration is calculated, and the change in grain size is analyzed within the grain growth trend range of 5 μm to 50 μm. This technical means simulates the dynamic evolution process of the grains through a thermo-mechanical coupling model, and finally obtains a cloud map of the grain size distribution. This cloud map shows the grain distribution during the rolling process, providing a basis for subsequent material property prediction, grain size control, and product quality optimization.

[0097] Preferably, step S22 includes the following:

[0098] Step S221: Perform an equivalent plastic strain increment dual-field analysis on the sheet thickness fluctuation to obtain sheet thickness-strain field information;

[0099] Step S222: Construct a thickness fluctuation constrained elastoplastic equation for the sheet thickness-strain field information and perform a derivation of the mixed drive constitutive parameters to obtain sheet mixed drive constitutive parameter data;

[0100] Step S223: Predict the rolling force for the sheet mixed drive constitutive parameter data to obtain the predicted value of the sheet rolling force.

[0101] In the embodiments of the present invention, through the equivalent plastic strain increment dual-field analysis of the sheet thickness fluctuation, the technical means realizes the linkage analysis of the thickness change and the strain field of the steel sheet. In this process, the plastic mechanics theory is adopted, combined with the method of incremental strain analysis, to model the mutual relationship between the sheet thickness and the strain field, and the sheet thickness-strain field data is obtained through numerical calculation. This data provides a basis for subsequent stress analysis, mechanical behavior prediction, and thickness control. Next, aiming at the characteristics of the sheet thickness-strain field, the elastoplastic equation construction of thickness fluctuation constraint is further carried out. By introducing the elastoplastic theory, the deformation behavior of the sheet is constrained under certain mechanical boundary conditions, providing a theoretical basis for establishing an accurate constitutive relationship. On this basis, the technical means quantitatively processes the mechanical properties of the material by deriving the constitutive parameters of hybrid drive and using specific physical and mechanical models, and finally obtains the constitutive parameter data of hybrid drive for the sheet. These parameters not only reflect the elastic and plastic characteristics of the material, but also combine the actual deformation conditions during the rolling process, providing a more accurate model support for the performance prediction of steel. Finally, the rolling force is predicted by using the Newton iteration method with thickness feedback for the constitutive parameter data of hybrid drive. This step uses the thickness feedback mechanism to adjust the control parameters during the rolling process in real time, and quickly converges to the appropriate rolling force prediction value through the Newton iteration method.

[0102] Preferably, step S23 includes the following:

[0103] Step S231: Extract the characteristics of the hydrostatic stress tensor according to the rolling stress tensor, and calculate the equivalent stress to obtain the rolling equivalent stress data;

[0104] Step S232: Evolve the dislocation density according to the rolling equivalent stress data to generate the rolling stress-driven evolution data; perform intelligent condition determination on the rolling stress-driven evolution data to obtain the recrystallization critical data.

[0105] In the embodiments of the present invention, the feature extraction of the rolling stress tensor is the starting point of this process. Through the feature extraction of the hydrostatic stress tensor of the rolling stress tensor, technical means are used to deeply analyze the external pressure and internal stress changes suffered by the material during the rolling process from the perspectives of mechanics and stress distribution. Subsequently, the equivalent stress data is further calculated, aiming to reflect the comprehensive stress state suffered by the material during the rolling process. This data plays a fundamental role in the subsequent micro-mechanical evolution and tissue change. Then, based on the rolling equivalent stress data, an algorithm for the evolution of dislocation density is used for calculation, so as to obtain the rolling stress-driven evolution data. This data represents the evolution process of the dislocation density in the material, which is an important feature of the plastic deformation of steel under high temperature and high pressure environments, and directly affects the microstructure and properties of the material. The analysis of the evolution of dislocation density provides an important basis for predicting the fatigue life, deformation mechanism, and stress concentration of the material. Finally, intelligent condition determination for recrystallization triggering is carried out with the help of the rolling stress-driven evolution data. This technical means aims to intelligently identify and determine the critical conditions for dynamic and static recrystallization based on real-time stress evolution data. Through in-depth analysis of the data, the system can identify the recrystallization process caused by internal stress in the material, thereby providing guidance for optimizing the rolling process and improving the material properties.

[0106] Preferably, the intelligent condition determination for recrystallization triggering includes the following:

[0107] When the real-time dislocation density ρ t is greater than or equal to ρ crit the intelligent condition determination is triggered; where ρ crit is calculated as follows:

[0108]

[0109] In the formula, ρ crit is the critical dislocation density, σ eq is the equivalent stress, α is the material constant, G is the shear modulus, b is the Burgers vector modulus, is the plastic strain rate, is the critical strain rate.

[0110] In the embodiments of the present invention, the mapping relationship between the α coefficient and the grain boundary characteristics is trained through in-situ electron backscatter diffraction (EBSD) data, and a high-throughput calculation model of the material constant is established; secondly, the spatial gradient distribution is captured in real time based on the distributed fiber optic strain sensor network, and the finite difference method is used to calculate the local strain rate tensor; finally, the its principal components, and dynamically correct the temperature dependence of the G value by combining the lattice distortion data measured by an online X-ray diffractometer. This multi-modal data fusion architecture effectively solves the prediction deviation caused by the assumption of ideal uniform deformation in traditional physical models, especially showing stronger robustness in the edge thinning area and the heterogeneous deformation area in the center of the rolled piece. By designing the parameters as functions of the process history (such as accumulated deformation work, cooling rate), the model further realizes the adaptive tracking of the dynamic-static recrystallization transformation boundary, providing a quantitative basis for the microstructure evolution for the real-time optimization of rolling process parameters.

[0111] Preferably, the obtained steel prediction digital twin model in step 2 includes the following:

[0112] Obtain a preset digital twin model;

[0113] Extract the austenite transformation rate of the steel according to the austenite transformation rate curve of the steel, and perform the treatment of the transformation of austenite in the steel into other phases to generate the austenite phase change parameters of the steel. Finally, inject them into the sub-module of the preset digital twin model for the prediction of the phase change process to obtain the phase change and grain evolution module;

[0114] Perform the analysis of the microstructure uniformity and local variation of the grain size distribution cloud map to generate the grain micro-feature data; perform the calculation of the real distribution of the grain micro-feature data, and finally perform the reconstruction of the micro-grain structure and generate the grain module to obtain the grain structure and distribution module;

[0115] Perform module fusion according to the phase change and grain evolution module and the grain structure and distribution module to generate a preliminary digital twin model; organically combine the predicted value of the sheet rolling force and the recrystallization critical data of the preliminary digital twin model, and perform model generation to obtain a preliminary fusion digital twin model;

[0116] Perform cross-module information integration on the preliminary fusion digital twin model and adjust the module parameters to obtain the steel prediction digital twin model.

[0117] In the embodiments of the present invention, by obtaining a preset digital twin model, a basic framework and a prediction platform are provided for the entire process. Then, based on the austenite transformation rate curve of the steel, the austenite transformation rate data of the steel is extracted therefrom, and the process of the austenite of the steel transforming into other phases is processed to generate the austenite phase change parameters of the steel. These phase change parameters are then injected into the phase change sub-module of the preset digital twin model for estimating the phase change process, and the result obtained is the phase change and grain evolution module. Next, for the grain size distribution cloud map, the system analyzes its microstructure uniformity and local variation to generate the grain microstructure characteristic data. These microstructure characteristic data are further calculated through the real distribution, and finally the reconstruction of the microscopic grain structure is realized to generate the grain structure and distribution module. In this process, the key to data processing lies in accurately capturing the distribution and variation of the grains and precisely reconstructing the grain structure through a physical model. After that, according to the phase change and grain evolution module and the grain structure and distribution module, module fusion is carried out, and the data of the two modules are comprehensively analyzed to obtain a preliminary digital twin model. Further, the system combines the predicted value of the plate rolling force and the recrystallization critical data, fuses the preliminary model data, and optimizes the accuracy of the steel process prediction to obtain a preliminary fusion digital twin model. Finally, through cross-module information integration and parameter adjustment, the preliminary fusion digital twin model is optimized to finally form a steel prediction digital twin model. In this process, the key technical means at the data level include phase change curve extraction, microstructure analysis, coupling and information sharing between modules, and dynamic optimization adjustment to achieve accurate prediction of the steel process.

[0118] As an example of the present invention, refer to Figure 2 shown, in this example, step S3 includes:

[0119] Step S31: Analyze the process parameter deviation degree of the steel prediction digital twin model by using the dynamic time warping algorithm to obtain the steel process deviation degree data;

[0120] Step S32: Perform data point matching and distance calculation according to the steel process deviation degree data to obtain the steel time series deviation degree map;

[0121] Step S33: Dynamically correct the tolerance boundary of the steel time series deviation degree map to obtain the dynamic correction result of the steel; formulate the defect risk level based on the dynamic correction result of the steel, and use the defect risk level to diagnose the steel process to obtain the steel process diagnosis result.

[0122] In the embodiments of the present invention, the Dynamic Time Warping (DTW) algorithm is used to analyze the process parameter deviation degree of the steel prediction digital twin model. By comparing the similarity between different time series, this algorithm can effectively identify the process parameter deviations of steel during the production process and generate corresponding process deviation degree data. This data reflects the changes at each time point in the steel process and provides a basis for subsequent analysis. Based on the obtained steel process deviation degree data, data point matching and distance calculation are further performed. In this process, the algorithm compares the process deviation degree data with historical data, calculates the similarity or difference degree between them, and then generates a time series deviation degree map of the steel. This map can present the process change trajectory of the steel during the entire production process in a time series manner, making any deviations or abnormalities in the production process visually presented and providing data support for subsequent processing. Further, dynamic correction of the tolerance boundary of the time series deviation degree map of the steel is carried out. The tolerance boundary is set according to historical process data and existing production requirements. As the production process progresses, the tolerance boundary needs to be dynamically adjusted according to the actual process fluctuations. This process compares the real-time process data with the set boundary values and makes appropriate corrections to ensure that the process parameters of steel production are within a reasonable range. The corrected data is used to formulate the defect risk level, and the steel process is diagnosed through the risk level, and finally the steel process diagnosis result is generated.

[0123] In this specification, an online prediction and diagnosis system for steel process data is provided, which is used to execute the above-mentioned online prediction and diagnosis method for steel process data. The online prediction and diagnosis system for steel process data includes:

[0124] A data acquisition and feature construction module deploys a distributed sensing network at key stations on the steel rolling production line to collect the rolling temperature gradient distribution, sheet thickness fluctuations, and roll pressure stress tensors, and constructs a multi-dimensional process feature data set;

[0125] A physical model coupling and digital twin model generation module performs microscopic physical constraints through the multi-dimensional process feature data set to generate multi-dimensional microscopic physical constraint data; decouples the multi-dimensional microscopic physical constraint data physically and data-wise, and constructs a digital twin model to obtain a steel prediction digital twin model;

[0126] A process deviation degree analysis and tolerance correction module analyzes the process parameter deviation degree according to the steel prediction digital twin model to generate steel process deviation degree data; dynamically corrects the tolerance boundary of the steel process deviation degree data to obtain a steel dynamic correction result; formulates a defect risk level based on the steel dynamic correction result, and diagnoses the steel process through the defect risk level to obtain a steel process diagnosis result;

[0127] The data push and diagnostic report generation module pushes the steel process diagnosis results to the cloud through the industrial Internet of Things platform for the closed-loop action of the diagnostic report, and obtains the online prediction and diagnostic report of the steel process.

[0128] The beneficial effects of the present invention are as follows: By deploying a distributed sensing network to collect various key process data (such as temperature gradient, sheet thickness fluctuation, and roll pressure stress tensor) during the rolling process in real time, and constructing a multi-dimensional feature data set, it provides accurate data support for process optimization and anomaly detection in the steel production process. Through the coupling of the dynamic heat conduction equation and the material phase transformation kinetics model, the austenite conversion rate and grain growth trend can be accurately calculated. Combining the elastic-plastic constitutive equation with the real-time stress-strain data, the rolling force prediction value is iteratively updated. This process, by integrating the dislocation density model and the recrystallization kinetics model, can not only predict the critical conditions of dynamic / static recrystallization, but also provides key data support for the construction of the digital twin model of steel, greatly enhancing the prediction ability of the steel production process. On this basis, the dynamic time warping algorithm (DTW) is used to analyze the process deviation degree of the steel prediction digital twin model, enabling the fluctuation of process parameters during the production process to be monitored in real time and compared with the standard process trajectory. The dynamic correction of the tolerance boundary ensures that the process deviation degree always remains within an acceptable range, and then generates accurate defect risk levels and process diagnostic reports. These diagnostic results are pushed to the central control system through the industrial Internet of Things platform, triggering the closed-loop action of the diagnostic report in real time, ensuring that the defect risks in the production process are timely feedback and processed, thereby guaranteeing the production quality and process stability. By integrating a distributed sensing network, physical model coupling, digital twin technology, and dynamic time warping algorithm, this system forms a comprehensive, accurate, and real-time responsive steel process diagnosis and optimization platform. It effectively solves the problems of lag in process anomaly detection, insufficient data analysis accuracy, and lack of real-time feedback mechanism in the traditional production process, improving the process control ability and product quality stability of steel production. Therefore, through the comprehensive application of distributed sensing technology, dynamic modeling, and real-time data analysis, the present invention solves the limitations of process deviation and defect prediction in traditional steel production, and improves the automation level and product consistency of the steel production process.

[0129] Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, it is intended to embrace all changes falling within the meaning and scope of the equivalent elements of the application documents within the present invention.

[0130] The above are only specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather will be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for online prediction and diagnosis of process data of steel, characterized in that: The following steps are involved: Step S1: deploy a distributed sensor network at key workstations of the steel rolling production line to collect rolling temperature gradient distribution, plate thickness fluctuation and roller stress tensor, and construct a multi-dimensional process feature data set; Step S2: Perform micro-physical constraints through a multi-dimensional process feature data set to generate multi-dimensional micro-physical constraint data; perform physical-data hybrid decoupling on the multi-dimensional micro-physical constraint data, and construct a steel prediction digital twin model; Step S3: Analyze the process parameter deviation according to the steel prediction digital twin model to generate steel process deviation data; Dynamically correct the tolerance boundary of steel process deviation data to obtain dynamic correction results of steel; Defect risk levels are formulated based on the dynamic correction results of steel materials, and process diagnosis of steel materials is performed based on the defect risk levels to obtain process diagnosis results of steel materials; Step S4: Push the steel process diagnosis results to the cloud through the industrial Internet of Things platform to perform a closed-loop diagnosis report to obtain an online predictive diagnosis report on the steel process.

2. The method for online prediction and diagnosis of process data of steel according to claim 1, characterized in that: The step S1 of constructing a multidimensional process feature data set includes: The multi-spectral infrared thermal imaging array deployed through the distributed sensor network monitors the gradient of 0-15mm below the surface of the rolled piece in the spectral range of 3-5μm and the sampling frequency of 200Hz to obtain the rolling temperature gradient distribution; According to the laser Doppler thickness measurement system, the initial inspection is carried out 10m at the exit of the finishing mill, and the final inspection is carried out 5m before the coiler. The scanning density is 256 points / section and the spacing is 8mm to obtain the plate thickness fluctuation; The piezoelectric sensor deployed in the six-dimensional force sensing roller system is used to collect the main component of the rolling force and the transverse stress component at a high speed of 2MHz and a wavelength demodulation frequency of 10kHz to obtain the roller pressure stress tensor; The data sets of rolling temperature gradient distribution, plate thickness fluctuation and roller stress tensor are integrated to obtain a multi-dimensional process characteristic data set.

3. The method for online prediction and diagnosis of process data of steel according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: plotting an austenite conversion rate curve for the rolling temperature gradient distribution to obtain an austenite conversion rate curve for the steel; and calculating a grain growth distribution diagram for the rolling temperature gradient distribution to obtain a grain size distribution cloud diagram; Step S22: using the elastic-plastic constitutive equation to predict the rolling prediction value of the plate thickness fluctuation to obtain the plate rolling force prediction value; Step S23: judging the critical recrystallization condition of the roller compression stress tensor to obtain critical recrystallization data of the steel; Step S24: Perform micro-physical customized convolution based on the steel austenite conversion rate curve, grain size distribution cloud map, plate rolling force prediction value and steel recrystallization critical data to generate multi-dimensional micro-physical constraint data; construct a steel prediction digital twin model based on the multi-dimensional micro-physical constraint data to obtain a steel prediction digital twin model.

4. The method for online prediction and diagnosis of process data of steel according to claim 1, characterized in that: Step S21 includes the following: Step S211: establishing a non-uniform temperature gradient tensor for the rolling temperature gradient distribution, and using an unstructured grid discretization process to obtain a three-dimensional temperature field of the steel; correlating the three-dimensional temperature field of the steel with the non-uniform temperature gradient tensor with alloy element concentrations in the range of 100-200 kJ / mol to obtain the activation energy of the steel phase change; Step S212: Calculate the austenite conversion rate according to the steel phase transformation activation energy to obtain the steel austenite conversion rate data; perform curve conversion on the steel austenite conversion rate data to generate the steel austenite conversion rate curve; Step S213: performing grain boundary migration calculation according to the non-uniform temperature gradient tensor, and analyzing the growth trend of grains between 5 μm and 50 μm to obtain a grain size distribution cloud map.

5. The method for online prediction and diagnosis of process data of steel according to claim 1, characterized in that: Step S22 includes the following: Step S221: performing an equivalent plastic strain increment double-field analysis on the plate thickness fluctuation to obtain the plate thickness-strain field information; Step S222: constructing a thickness fluctuation constrained elastoplastic equation for the plate thickness-strain field information, and deriving hybrid drive constitutive parameters to obtain hybrid drive constitutive parameter data for the plate; Step S223: performing rolling force prediction on the plate hybrid drive constitutive parameter data to obtain a predicted value of the plate rolling force.

6. The method for online prediction and diagnosis of process data of steel according to claim 1, characterized in that: Step S23 includes the following: Step S231: extracting the hydrostatic stress tensor feature according to the roller pressure stress tensor, and calculating the equivalent stress to obtain roller pressure equivalent stress data; Step S232: performing dislocation density evolution according to rolling equivalent stress data to generate rolling stress driven evolution data; The recrystallization triggering intelligent condition judgment is performed on the roller pressure stress driven evolution data to obtain the recrystallization critical data.

7. The method for online prediction and diagnosis of process data of steel according to claim 6, characterized in that: The recrystallization triggering intelligent condition determination includes: When the real-time dislocation density ρ t Greater than or equal to ρ crit Trigger intelligent condition determination; where ρ crit The calculation is as follows: In the formula, ρ crit is the critical dislocation density, σ eq is the equivalent stress, α is the material constant, G is the shear modulus, b is the Burgers vector modulus, is the plastic strain rate, is the critical strain rate.

8. The method for online prediction and diagnosis of process data of steel according to claim 1, characterized in that: The construction of the steel prediction digital twin model described in step 2 includes the following: Get the preset digital twin model; The steel austenite conversion rate is extracted according to the steel austenite conversion rate curve, and the austenite in the steel is transformed into other phases to generate the steel austenite phase transformation parameters. Finally, the preset digital twin model submodule is injected to estimate the phase transformation process to obtain the phase transformation and grain evolution module. Perform microstructure uniformity and local variation analysis on the grain size distribution cloud map to generate grain microscopic feature data; perform true distribution calculation on the grain microscopic feature data, and finally reconstruct the microscopic grain structure and generate the grain module to obtain the grain structure and distribution module; The modules of phase change and grain evolution module and grain structure and distribution module are integrated to generate a preliminary digital twin model; the plate rolling force prediction value and recrystallization critical data are organically combined in the preliminary digital twin model, and the model is generated to obtain a preliminary integrated digital twin model; The preliminary fusion digital twin model is integrated across modules and the module parameters are adjusted to obtain the steel prediction digital twin model.

9. The method for online prediction and diagnosis of process data of steel according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: using a dynamic time warping algorithm to perform process parameter deviation analysis on the steel prediction digital twin model to obtain steel process deviation data; Step S32: performing data point matching and distance calculation according to the steel process deviation data to obtain a steel time series deviation map; Step S33: dynamically correct the tolerance boundary of the steel material time series deviation map to obtain the dynamic correction result of the steel material; formulate the defect risk level based on the dynamic correction result of the steel material, and use the defect risk level to perform process diagnosis on the steel material to obtain the steel material process diagnosis result.

10. An online prediction and diagnosis system for process data of steel, characterized in that: The method for online prediction and diagnosis of process data of steel materials according to claim 1 is used to implement the method, wherein the online prediction and diagnosis system for process data of steel materials comprises: Data collection and feature construction module, deploys distributed sensor networks at key stations of steel rolling production lines, collects rolling temperature gradient distribution, plate thickness fluctuation and roller stress tensor, and constructs a multi-dimensional process feature data set; The physical model coupling and digital twin model generation module performs microscopic physical constraints through multidimensional process feature data sets to generate multidimensional microscopic physical constraint data; performs physical-data hybrid decoupling on the multidimensional microscopic physical constraint data, and constructs a digital twin model to obtain a steel prediction digital twin model; The process deviation analysis and tolerance correction module analyzes the process parameter deviation according to the steel prediction digital twin model and generates steel process deviation data; dynamically corrects the tolerance boundary of the steel process deviation data to obtain the dynamic correction result of the steel; formulates the defect risk level based on the dynamic correction result of the steel, and performs process diagnosis on the steel according to the defect risk level to obtain the steel process diagnosis result; The data push and diagnostic report generation module pushes the steel process diagnostic results to the cloud through the industrial Internet of Things platform to complete the diagnostic report closed-loop action and obtain an online predictive diagnostic report on the steel process.

Images