Efficient intelligent monitoring and digitalization technology for continuous casting
1. Research Background
Continuous casting is the core process that links steelmaking and steel rolling. Its stability and intelligence level directly affect product quality, production efficiency, and equipment safety. Monitoring of the process, especially real-time tracking of the solidification process of the casting flow, monitoring of surface and internal deformation behaviors, and prediction of the dynamic evolution of the microstructure, not only affects the quality of the cast billet, but also has a significant impact on the performance regulation of the subsequent rolling process and the performance of the final rolled product. At the same time, faults in the casting rolls, abnormalities in the secondary cooling nozzles, etc., have a significant impact on the quality of the cast billet. It is urgent to use data-driven diagnostic models for real-time warnings. This study focuses on the metallurgical behavior and key equipment status of the continuous casting process. By integrating mechanism modeling and data-driven methods, a process simulation, intelligent monitoring, and process control model system integrating solidification tracking, deformation prediction, microstructure evolution, and equipment anomaly diagnosis has been constructed. A complete closed loop of "state perception - data transmission - model fusion - monitoring simulation - optimization control" has been established to ensure the stability and reliability of the continuous casting product quality (Figure 1).
2 Key Technologies Introduction
2.1 Prediction Technology for Continuous Casting Billet Solidification Process
The prediction technology of the solidification process of the billet is one of the core technologies of the digital twin system for continuous casting. The breakthrough points of this technology include the dynamic characterization of thermal physical parameters based on the microscopic segregation model, the coupling of multiphase solidification model and solute evolution law, the processing of non-uniform cooling boundary conditions, and the mechanism of the influence of the reduction process on the solidification process. The core of the technology lies in constructing a dynamic thermal physical parameter library related to the composition, temperature and solute segregation of the steel grade, achieving high-precision real-time prediction of the solidification process of the billet.
2.2 Prediction Technology for Cast Billet Deformation Characteristics
The casting billet deformation characteristic prediction technology combines deep learning with numerical simulation to achieve rapid and accurate prediction of the stress-strain distribution during the upsetting process at the end of solidification of the casting billet. The core of the technology includes a rapid derivation model for strain distribution based on the self-attention-regulated production adversarial network (SACGAN), a multi-modal deep learning rheological curve prediction model based on the encoder-decoder framework, and an accurate calculation method for the pressed billet force. This technology solves the technical problem of long calculation time of traditional finite element simulation and the difficulty in meeting the requirements of online applications.
2.3 Prediction Technology for Cast Billet Structure Evolution
The technology for predicting the microstructure evolution of cast billets mainly addresses the accurate characterization of the second phase precipitation and continuous phase transformation during the continuous casting solidification process. This is a key technology that affects the thermal plasticity and enables the integrated regulation of the casting-rolling microstructure. The breakthrough lies in using the micro-element superposition method to characterize the growth of the second phase under varying cooling rates and predicting the phase transformation process based on the "mixed model + solute resistance theory". The core of the technology is to consider the influence of solute segregation and the difference in cooling rates at different positions of the cast billet's cross-section, and establish a kinetic model for the second phase precipitation under varying cooling rates and a prediction model for the austenite-ferrite phase transformation, achieving accurate prediction of the microstructure evolution characteristics.
2. Four-roll gap anomaly diagnosis technology
The continuous casting roll gap anomaly diagnosis technology aims to address the risks of missed inspections and incorrect judgments in traditional manual inspections. It achieves precise monitoring and fault warning of the roll gap status through intelligent means. The key technological breakthroughs include the processing of industrial big data and the definition of abnormal characteristics of the roll gap. The core lies in the effective integration of multiple sources of industrial big data and the precise extraction of fault features. The core of this technology is to construct a multi-dimensional relationship model of "process - pressure - casting force - roll gap", combined with metallurgical expert knowledge and data mining algorithms, to accurately identify typical faults such as deviation in the connection of the fan-shaped section and asymmetry of the roll gap on the left and right.
2.5 Abnormal Nozzle Diagnosis Technology
The abnormal diagnosis technology for the secondary cooling circuit nozzles has solved the key problems existing in traditional manual inspection or spray testing, such as long monitoring periods, low positioning accuracy, and inability to provide real-time warnings. This technology avoids the limitations of the indirect assessment method based on the surface temperature of the cast billet, which is susceptible to interference from factors such as water vapor and iron oxide scale. Instead, it directly establishes a nozzle fault diagnosis model based on industrial big data from the production process.
15358968703