The steel production process is a large and complex industrial process covering multiple processes and multiple levels of control. Most steel companies in our country are equipped with fully functional one to five-level control systems and have a high level of automation, but there are still “two-dimensional problems” horizontally and vertically. Therefore, the intelligent manufacturing idea of “two-dimensional strategy” has been widely recognized, and companies can Promote the intelligent application process of the steel industry system from both horizontal and vertical dimensions.
Vertical application refers to the “end-to-end information fusion” of the five-level systems within the enterprise, achieving seamless connections from the lowest-level drivers and actuator signals to the highest-level enterprise asset management system. Horizontal applications mainly refer to the integration of information and services between steel companies and upstream and downstream of the industry chain, to achieve value chain integration and collaborative optimization of the entire steel industry chain, and to provide intelligent solutions for the entire value chain of the entire industry. The core of realizing this “two-dimensional strategy” is CPS intelligent technology based on digital perception. The CPS intelligent key technology based on digital perception, will be a major strategic direction for the development of the steel industry in the past ten years to achieve multi-process, system-level and global-level product quality and production process optimization in the entire steel production process.
1. Blast furnace big data modelling and intelligent optimization
Build a front-end blast furnace big data collection and transmission system through the Internet of Things, and perform big data processing and analysis through the cloud platform, effectively improving corporate management efficiency and production process optimization, and promoting traditional steel companies to become intelligent. Without affecting the existing systems and application architecture of the steel plant, an enterprise-wide blast furnace big data privatized cloud service system was built by integrating the existing blast furnace information system. Through the design and development of interactive functions of the big data cloud platform, blast furnaces big data collection, data processing, data storage, and client interaction are realized.
The edge computer completes the real-time data collection and data preprocessing of blast furnace production and deploys the data correlation analysis and mining system, blast furnace mechanism modelling and digital simulation system, blast furnace process analysis and prediction, furnace condition comprehensive evaluation system, and foreman support in the blast furnace private cloud. system, data-based autonomous learning system and blast furnace visualization system.
2. Intelligent model and optimization of steelmaking process
The modern converter steelmaking process includes molten iron pretreatment → top and bottom blowing converter steelmaking → refining outside the furnace (RH vacuum refining, LF refining) → continuous casting and other major production processes. Molten iron pretreatment, converter steelmaking blowing and out-of-furnace refining are all complex black box systems with dynamic changes in the multi-phase reaction of slag-gold-gas, involving the interaction of many parameters such as the composition and temperature of the high-temperature melt, and the operating process. These influencing factors are strongly coupled, multi-variable, and nonlinear, making it very difficult to accurately control the molten iron pretreatment, converter steelmaking, and out-of-furnace refining processes.
Steelmaking process control has gone through five stages: empirical control, static control, dynamic control, fully automatic control, and intelligent control. The shortcomings of traditional steelmaking automation mathematical models such as traditional mechanism models, incremental models, neural network models, empirical models, statistical models, expert systems, etc. are poor adaptability, accuracy that needs to be further improved, low intelligence, and can only be used locally. Optimization cannot be used for high-precision dynamic coordinated control and global optimization.
Use “industrial big data + artificial intelligence + metallurgical mechanism” to establish an intelligent mathematical model and develop a digital twin system that fully corresponds to the physical entities of molten iron pretreatment, converter blowing, and refining outside the furnace and has the capabilities of perception, analysis, decision-making, and execution. It can truly reflect the behaviour and status of the actual steelmaking process, enhance the adaptability and intelligence level of the model, realize intelligent optimization control of molten iron pretreatment, converter blowing, and out-of-furnace refining, improve the control hit rate, and provide a basis for the steelmaking system. Intelligent manufacturing and the implementation of intelligent manufacturing in the entire production line from ironmaking, steelmaking, and continuous casting to steel rolling have laid the foundation. In terms of molten iron pretreatment, converter steelmaking, and out-of-furnace refining, the purpose of reducing steelmaking material and energy consumption, improving molten steel quality, improving production efficiency, and reducing production costs can be achieved.
3. Continuous casting big data modelling and process monitoring
The continuous casting billet quality analysis and forecasting system is a computer judgment system designed to solve the billet quality problems in continuous casting production. Apply artificial intelligence technology, based on the knowledge and experience provided by many experts in the field of continuous casting, to simulate the expert decision-making process and solve complex problems related to slab quality control.
On the basis of analyzing the development and application of continuous casting billet quality analysis and forecasting systems at home and abroad, combined with the characteristics of my country’s continuous casting production and the main quality defects, through the feature extraction and classification of continuous casting field data, we can analyze the casting The main quality defects existing in the blank are obtained, and the probability of defect occurrence and the stepwise distribution of each process parameter and quality defect level are obtained at the same time; through the establishment and analysis of the quality defect fault tree, the probability of defect occurrence and the influence of each process parameter on quality defects are obtained. influence level. According to the model analysis characteristics, the objective function of quality analysis and prediction is proposed, a continuous casting billet quality analysis and prediction model is established based on the field data analysis model and the fault tree statistical model, and the corresponding analysis and prediction plan is formulated.
4. Intelligent optimization of the rolling process
Artificial intelligence technology has been applied to many aspects of the steel rolling field. Neural networks are used to predict rolling force, identify roll eccentricity, shape control, comprehensive control of shape and thickness, and predict the microstructure and properties of hot-rolled strips. In order to improve the rolling force presetting accuracy of the finishing rolling unit, the information processing tool of the neural network is used in the strip rolling setting calculation. After using this system, the rolling force prediction accuracy is significantly improved.
Fuzzy logic and fuzzy control are used for the allocation of continuous rolling schedules, cold-rolled strip shape control, fuzzy dynamic setting of hot-rolled strip head rolling, segmented roll cooling, and plate thickness-tension fuzzy decoupling control. Fuzzy control is introduced into the flatness control of cold tandem rolling. Actual operation results show that after using fuzzy control, the shape deviation is reduced by half, high-speed rolling can be achieved, and production efficiency is significantly improved.
As an important branch of artificial intelligence, the expert system was initially used for production processes that did not require high real-time requirements, such as heating furnace discharge rhythm control, rolling load distribution, coil transportation on the finishing line, and strip thickness accuracy diagnosis. Used for diagnosis, control, planning and design, logistics management systems, etc. In recent years, with the gradual improvement of expert system theory and the rapid development of computer technology, expert systems have begun to be used in some real-time control systems such as rolling schedule setting and control and flatness control, and have achieved good results. In addition to the above methods, another new technology of artificial intelligence, the genetic algorithm, has also begun to be applied in the steel rolling process. The genetic algorithm is used to optimize the rolling schedule and predict the short-stroke control parameters of the vertical roller.
At present, the development trend of artificial intelligence technology in the steel rolling industry is to make full use of the advantages of various artificial intelligence technologies and combine them to overcome the inherent shortcomings of various artificial intelligence technologies and minimize Human intervention in the system to maximize the control system accuracy of the rolling process.
5. Intelligent maintenance of process equipment
Intelligent maintenance changes the traditional passive maintenance model into a proactive maintenance model, using the development and application of big data-driven information analysis, performance degradation process prediction, maintenance optimization, on-demand monitoring, etc. Equipment maintenance reflects preventive requirements. This results in near-zero failures. Furthermore, the equipment intelligent monitoring model based on big data-driven fault prediction predicts key equipment life measurement functions, health factors and other performance variables, and establishes a health management system including fault cause and effect diagrams, statistical analysis, remaining life estimation, and maintenance plan decisions. , conduct a comprehensive evaluation based on actual conditions such as equipment health indicators, order status, and product quality requirements within the process, and provide the best process equipment maintenance plan to achieve predictive maintenance and reduce downtime.
6. Conclusion
For the entire production process of ironmaking, steelmaking, continuous casting, rolling and post-rolling treatment, a steel process quality big data platform is constructed to study key parameter detection technologies such as surface quality, three-dimensional morphology, molten steel temperature, and structural properties of each process. Real-time analysis of the relationship between production process technology, equipment parameters and product quality, improving mathematical model setting and quality control accuracy based on data, and improving product quality stability through multi-process coordination and matching. Develop fault diagnosis and coordination optimization technology for key equipment to form a remote monitoring and fault diagnosis system platform. Break through process interface and system barriers, realize horizontal, vertical and end-to-end integration of the steel industry, and build a reliable, real-time, collaborative steel intelligent CPS system.