The importance of clean steel to product quality is increasingly being recognized. Controlling the size distribution, morphology, and composition of the non-metallic oxide inclusions is the first demand for clean steel. In addition, sulfur, phosphorus, hydrogen, nitrogen and even carbon [1, 2] also should be controlled, because these elements affect steel mechanical properties. For example, formability, ductility and fatigue strength worsen with increasing sulfide and oxide inclusion content. Lowering C and N enhances strain aging and increases ductility and toughness. Solid solubility, hardenability and resistance to temper embrittlement can be enhanced by lowering P. [1] The definition of ‘clean steel’ varies with steel grade and its end use. For example, IF steel requires C and N both <30ppm; line pipe requires S, N and O all <30ppm; HIC resistant steel requires P≤50ppm and S≤10ppm, and bearing steel requires the total oxygen less than 10ppm. In addition, many applications restrict the maximum size of inclusions [3, , so the size distribution of inclusions is also important. The control of steel cleanliness has been extensively reviewed by Kiessling in 1980 , McPherson and McLean in 1992 , Mu and Holappa in 1993 , Cramb in 1999 , and Zhang and Thomas in 2003 . The current paper reports on steel cleanliness investigations and mathematical simulations of fluid flow and inclusion behavior in ladle, tundish and continuous caster to improve understanding of clean steel production.