International Conference on Automation and Engineering (AUEN 2022)
March 12~13, 2022, Virtual Conference
Long Short-Term Memory Neural Networks for Clogging Detection in the Submerged Entry Nozzle
Ana P. M. Diniz, Patrick M. Ciarelli, Evandro O. T. Salles and Klaus F. Coco, Universidade Federal do Espírito Santo, Vitória, Espírito Santo, Brazil
The clogging in the Submerged Entry Nozzle (SEN), responsible for controlling the steel ow in continuous casting, is one of the major problems faced by steelmaking process, since it can increase the frequency of interruptions in the operation for the maintenance and/or exchange of its equipment. Although it is a problem inherent to the process, not identifying the clogging can result in losses associated with the process yield, as well as compromising the product quality. In order to detect the occurrences of clogging in a real steel industry from historical data of process variables, in this paper, different models of Long Short-Term Memory (LSTM) neural networks were tested and discussed. The overall performance of the classifiers developed here showed very promising results in real data with class imbalance.
Continuous Casting, Submerged Entry Nozzle, Clogging, LSTM, Deep Learning.