An efficient computational technique for the prediction of wire rod temperatures under different industrial process conditions

Abstract

Mechanical properties of industrial manufactured wire rod depend on the chemical composition of the steel and the air-cooling rates underwent by the product after hot rolling, generated by fans located beneath a conveyor. The number of operational variables that determines the cooling process, given by the conveyor velocity and the different fan sections powers (controlled independently), lead to a dependency of the cooling on a high multidimensional parameter space whose potential combinations are impossible to be analyzed, either experimentally or by numerical simulation of a thermal-metallurgical model. To tackle this problem, an efficient strategy, based on the use of Higher Order Singular Value Decomposition (HOSVD), is presented. This approach consists firstly in the prediction of the cooling rate using numerical simulation of a thermal-metallurgical model for a reduced number of points of the parameter space and therefore the generation of a database with these simulated cases. Then, the HOSVD technique is applied to provide simplified global descriptions of the multidimensional database that, finally, permits to predict quite accurately the cooling curve for any combination of the process parameters. The work presented shows a robust and simple strategy on the selection of (a reduce number of) particular combinations of process parameters that are used to construct the full database. Fast on-line predictions of the cooling rates (once the database has been built) allow to incorporate accurate modelling results in many Engineering tools, such as model predictive control algorithms or plant simulation software.