Cracks Prediction using Artificial Neural Networks in Continuous Casting of Steel

G.O. Tirian, P.B. Camelia, and S. Rusu-Anghel (Romania)


Prediction, neural networks, cracks, continuous casting.


In this paper, we introduce a new multi-neuronal system that allows us to regulate the speed of the wire drawing, based on cracks prediction. When the segment that cracks gets out of the crystallizer, the melted steel pours out and the casting process should be stopped. We should avoid such incidents by detecting the cracks and reducing the speed of the wire drawing which allows the steel to get solid. It is established then when a crack occurs, the liquid steel touches the crystallizer’s wall, causing an increase in its temperature. This suggests that cracks may be detected by means of several heat sensors mounted on the crystallizer’s wall both on its width and on the direction of casting. An artificial neuronal network could analyse all the data received by the sensors mounted on the walls of the crystallizing apparatus and could acknowledge any crack within a large range of precision.

Important Links:

Go Back