Application of k-means clustering for temperature timing characteristics in breakout prediction during continuous casting

Abstract

Breakout is the most serious incident in continuous casting. The missing and false alarms will seriously damage the caster and greatly affect the quality of slabs. In view of the defects of existing breakout prediction methods, k-means clustering and dynamic time warping (DTW) are combined to investigate and develop an effectual prediction method. Through extracting the typical temperature timing characteristics from the temporal and spatial perspectives, the method based on k-means clustering and DTW is proposed to distinguish and recognize the breakout. Compared with in-service breakout prediction system (BPS), the prediction results of the proposed method confirm that the number of false alarms can be reduced from 50 to 8 while ensuring a 100% correct alarm rate. The excellent prediction performance demonstrates that the clustering-based breakout prediction method exhibits good application potential, while it offers a novel approach to monitoring abnormalities in the continuous casting process.