The development of high performance steels grades conducts to more and more severe requirements in terms of inclusion cleanness. Steelmakers need to control very early the inclusions present at the liquid state in order to improve the process and achieve the final product quality.
Previous studies have shown that the inclusions observed in the samples taken at the liquid state are not always the inclusions really present in the liquid steel. Taking into account that a wrong cleanness evaluation can conduct to quality troubles and to customers' claims the main objective of this project was to determine the most reliable liquid steel sampling method and the most appropriate inclusion assessment strategy.
The second objective was the prediction of the final product cleanness on the basis of liquid steel analysis.
By coupling modelling, laboratory scale experiment, industrial sampling campaigns, statistical approaches and improvement of certain techniques, following conclusions have been obtained:
- the most reliable sampling procedure is to use an argon blowing device which avoids slag entrapments, steel reoxidation and un-controlled shrinkage location. The lollipop sample shapes are recommended;
- the most reliable characterisation strategy is to implement PDA which gives an oxide cleanness index, then automatised SEM (oxides: 2 μm - 10 μm) and, if necessary, automatised LOM (oxides > 10 μm) and EE (inclusions < 2 μm);
- the prediction of the oxide composition in the final product on the basis of the tundish samples is possible. The prediction of density and size is more difficult because it depends on inclusion deformation during rolling.