Título

EVALUATION OF CAS-OB PROCESS CONDITIONS IN THE OCCURRENCE OF LADLE NOZZLE CLOGGING THROUGH MACHINE LEARNING

EVALUATION OF CAS-OB PROCESS CONDITIONS IN THE OCCURRENCE OF LADLE NOZZLE CLOGGING THROUGH MACHINE LEARNING

Autoria

Melo, Pedro Henrique Resende Vaz de; Silva, Marlon José Dos Anjos; Facundes, Willian; Bielefeldt, Wagner Viana

DOI

10.5151/2594-5300-39452

Downloads

Baixar Artigo 22 Downloads

Resumo

Ladle nozzle clogging is a recurring problem in steel shops worldwide. This event may cause heat return to BOF, interruptions in the continuous casting sequence, decrease continuous casting speed, and remove the ladle from the production cycle, leading to productivity losses. The causes of clogging are generally associated with the deposition of non-metallic inclusions (NMI) on valve walls or the solidification of steel by low temperature. Several factors can affect the behavior of NMI during the process, from slag chemical composition, deoxidation, calcium treatment, and flotation time. Thus, based on an exploratory analysis of a database of industrial heats, this study aimed to determine which process parameters have the highest correlation with clogging events in steels treated in the CAS-OB route. Machine learning algorithms were used to select important variables and develop a model to classify clogging events. The model achieved a classification performance of 66% and was explained using the Shapley values method, which considered the influence of calcium content, valve life, desulfurizer weight and NMI removal mechanisms such as the use of argon lance and porous plug. Based on these results, it was possible to propose actions to reduce the incidence of ladle nozzle clogging

 

Ladle nozzle clogging is a recurring problem in steel shops worldwide. This event may cause heat return to BOF, interruptions in the continuous casting sequence, decrease continuous casting speed, and remove the ladle from the production cycle, leading to productivity losses. The causes of clogging are generally associated with the deposition of non-metallic inclusions (NMI) on valve walls or the solidification of steel by low temperature. Several factors can affect the behavior of NMI during the process, from slag chemical composition, deoxidation, calcium treatment, and flotation time. Thus, based on an exploratory analysis of a database of industrial heats, this study aimed to determine which process parameters have the highest correlation with clogging events in steels treated in the CAS-OB route. Machine learning algorithms were used to select important variables and develop a model to classify clogging events. The model achieved a classification performance of 66% and was explained using the Shapley values method, which considered the influence of calcium content, valve life, desulfurizer weight and NMI removal mechanisms such as the use of argon lance and porous plug. Based on these results, it was possible to propose actions to reduce the incidence of ladle nozzle clogging

Palavras-chave

Ladle nozzle clogging; Non-metallic inclusions; Secondary refining, Machine learning

Ladle nozzle clogging; Non-metallic inclusions; Secondary refining, Machine learning