Abstract
A methodology for building pragmatic physics-based models is here adapted to predict the erosion of ladle linings in the steel industry, in order to support operators when deciding whether the lining can be used safely for another heat. A defective lining may allow 140 t of molten steel to spill, with disastrous consequences for workers and plant. The adopted work flow for the development,
challenges faced, and some model results are presented. One key learning outcome is that model development should allow time for maturing the process understanding, as well as for many iterations by ‘questions-responses and actions’ at various stages in the model development. Good interaction between the development team and industry case owner is an important success factor. Combining or extending the model with the use of machine learning and cognition-related methods, such as knowledge graphs and self-adaptive algorithms, is discussed.
challenges faced, and some model results are presented. One key learning outcome is that model development should allow time for maturing the process understanding, as well as for many iterations by ‘questions-responses and actions’ at various stages in the model development. Good interaction between the development team and industry case owner is an important success factor. Combining or extending the model with the use of machine learning and cognition-related methods, such as knowledge graphs and self-adaptive algorithms, is discussed.