Continuous casting is a dominant steelmaking process due to its steady-state nature, enhanced yield of steel production, and low consumption of energy and manpower. However, the production of defect-free cast products is still a big challenge, as the internal defects, including macro-segregation of alloying elements, cracks, and pores, can be easily formed during the solidification process, which seriously deteriorates the microstructure and mechanical properties of the cast products. Therefore, this paper puts forward the common solutions to overcome these problematic issues. The reduction technology can effectively improve the center segregation of the billet, bloom, and slab. The history of the development of the reduction technology is summarized and classified according to the variations in research trend. Furthermore, the basic principles and parameters of reduction technology are described to implement them in the actual production process. This paper compares the similarities and differences between soft reduction (SR) and heavy reduction (HR) technologies with a particular focus on the theoretical research of HR technology and further elaborates the key parameter and equipment problems during implementation of HR. Moreover, this paper also considers the HR technology adopted by Baosteel as a case study, which helped to put forward some viewpoints for the future development of reduction technology.

With the improvement of computer computing power and the development of big data technology, neural networks have rapidly developed and been effectively applied in multiple fields. Steelmaking and continuous casting are key stages in steel production, involving complex physical changes and chemical reactions under high-temperature conditions. Neural networks, with the advantage of intelligently mining patterns from massive data, have been extensively researched and applied in the control of steelmaking and continuous casting processes. This paper uses a systematic literature review method to summarise the applications of neural networks in converter steelmaking, electric arc furnace (EAF) steelmaking, secondary refining and continuous casting over the past two decades. The main areas include the prediction of endpoint temperature and composition in converter and EAF smelting, optimisation of EAF power supply parameters, prediction of breakout in continuous casting and predictive control of billet and slab defects in continuous casting. It summarises and analyses the advantages and effects of different types of neural networks applied in steelmaking and continuous casting. Finally, the paper highlights the potential and challenges of neural networks in the application of steelmaking and continuous casting and points out future development directions.