Blom/etal/2014a: Reliable BOF endpoint prediction by novel data-driven modeling

Bibtype Inproceedings
Bibkey Blom/etal/2014a
Author Schlüter, Jochen and Odenthal, Hans-Jürgen and Uebber, Norbert and Blom, Hendrik and Beckers, Tobias and Morik, Katharina
Ls8autor Beckers, Tobias
Blom, Hendrik
Morik, Katharina
Title Reliable BOF endpoint prediction by novel data-driven modeling
Booktitle AISTech Conference Proceedings
Publisher AISTech
Abstract In a cooperative effort between SMS Siemag AG (SMS), AG der Dillinger Hüttenwerke (DH), and the Chair for Artificial Intelligenceof the TU Dortmund University (TUD), modern data mining and machine learning algorithms have been closely integrated with steelprocessing by a newly developed modular, distributed and scalable data-stream processing system. It can be easily extended to readand analyze data from all kinds of sensors and data sources. Its analysis can readily be applied to metallurgical and manufacturingprocesses.The new approach makes use of a variety of static and dynamic process and measurement variables and uses the precision of data mining and its adaptability to the 190 t converters of DH in order to predict target variables, such as the temperature or the phosphoruscontent of the melt at the end of blowing.Rough physical conditions can lead to sensor failures and deterioration. Use is made of statistical methods and multiple models basedon different process in order to cope with these technical issues.
Year 2014