Slag Recognition in Steel Plant Converters Tapping

The correct identification of the moment to end tapping converter in a steel plant represents a significant cost reduction. Once the slag is a by-product generated within the converter during the blowing process, it is not desired that it be carried to the following processes. From the images generated by a thermographic camera, digital image processing techniques will be applied to identify and segment the region of steel jet formed during the converter tapping to the ladle steel. In addition, feature vectors from segmented jet images of 97 heats will be extracted and pattern recognition techniques will be applied to classify the slag carry over in this process, determining the end of tapping. The SVM, kNN, ANN and CUSUM methods will be employed using the generated characteristic vectors. Their results demonstrated that the applied techniques can reach results close to 100% of the correct classification of the slag carry over.