Abstract: In the steel industry, the ironmaking process involves complex and high-risk environments with extreme temperatures and high pressures, exposing frontline workers to high safety risks for long time. Against the background of smart manufacturing and industrial digital transformation, building an intelligent safety monitoring system for the ironmaking process has become one of the core paths for the high-quality development of the industry. This paper systematically reviews the evolution of ironmaking safety monitoring from mandatory management to intelligent early warning, and analyzes the research progress and application practices of three core dimensions including personnel, equipment and process based on the "Human-Machine-Material-Environment-Process" framework. Significant breakthroughs have been made in multi-source information perception, intelligent risk identification and cross-system coordinated early warning. Personnel safety can realize quantitative early warning of typical risks such as irregular operation and on-the-job fatigue through the integration of heterogeneous positioning, visual recognition and physiological state monitoring. Equipment safety initially constructs a full-life-cycle fault prediction and health management (PHM) system for key equipment based on multi-source state perception and hybrid intelligent models; process safety significantly extends the early warning window for abnormal furnace conditions by means of global perception and data fusion technologies, and some systems have realized cross-process coordinated control. Combined with the engineering practices of enterprises such as Wuhan Iron and Steel Co., Ltd. and Liuzhou Iron and Steel Co., Ltd., the intelligent safety monitoring system has achieved quantifiable benefits in reducing blast furnace downtime, improving labor productivity and optimizing energy consumption. However, it still faces bottlenecks in large-scale deployment, such as data silos, insufficient model generalization ability, high construction and operation costs, shortage of interdisciplinary talents and lack of unified standards. For the future, to build an intrinsically safe, intelligent and efficient ironmaking production environment, it is necessary to further develop hybrid-driven models embedded with metallurgical mechanisms to improve model interpretability and working condition adaptability. At the same time, a safety-oriented unified data platform should be built to break information barriers and support cross-system coordinated early warning and regulation.