Control of non-metallic inclusions is essential for the production of high-quality steel. This review summarizes processes that change inclusion compositions and concentrations during secondary steelmaking—slower changes are limited by reaction between bulk steel and slag or refractory, and faster changes involve direct additions to the steel bath. An example of the former is conversion of alumina inclusions to spinels during ladle treatment, while reoxidation and calcium treatment are typical exemplars of the fast changes. For the slower changes, inclusions approach equilibrium with the liquid steel and conceptually simple kinetic models correctly describe inclusion evolution during ladle treatment. Disequilibrium from faster changes persists for several minutes under typical ladle conditions, with small-scale inhomogeneity in the steel. Fast scanning electron microscopy with microanalysis has facilitated detailed study of these inclusion evolution processes by providing information on inclusion composition, size, and shape. Machine learning methods are likely to be increasingly important in analysis of the results. Such methods have already shown promise to improve classification of inclusions and recognizing inclusion clusters, from analyses of polished sections. Several unresolved issues that require future study are noted.

中文翻译：

---

钢材加工对氧化物清洁度的考虑回顾

控制非金属夹杂物对于生产优质钢至关重要。本综述总结了在二次炼钢过程中改变夹杂物成分和浓度的过程——较慢的变化受到大宗钢与炉渣或耐火材料之间的反应的限制，而较快的变化涉及直接添加到钢浴中。前者的一个例子是在钢包处理过程中氧化铝夹杂物转化为尖晶石，而再氧化和钙处理是快速变化的典型例子。对于较慢的变化，夹杂物与钢水接近平衡，概念上简单的动力学模型正确地描述了钢包处理过程中夹杂物的演变。在典型的钢包条件下，快速变化的不平衡会持续几分钟，钢中存在小规模的不均匀性。带微量分析的快速扫描电子显微镜通过提供有关夹杂物成分、大小和形状的信息，促进了对这些夹杂物演化过程的详细研究。机器学习方法在结果分析中可能变得越来越重要。通过对抛光切片的分析，这些方法已经显示出改善夹杂物分类和识别夹杂物簇的前景。指出了需要未来研究的几个未解决的问题。通过对抛光切片的分析，这些方法已经显示出改善夹杂物分类和识别夹杂物簇的前景。指出了需要未来研究的几个未解决的问题。通过对抛光切片的分析，这些方法已经显示出改善夹杂物分类和识别夹杂物簇的前景。指出了需要未来研究的几个未解决的问题。