Reboud, Christophe; CEA LIST; France
Skarlatos, A.; CEA; France
Reboud, C.; CEA, LIST; France
Kebe, T.; ThyssenKrupp Steel Europe; Germany
van den Berg, F.D.; Tata Steel; Netherlands
Martinez-De-Guerenu, A.; CEIT; Spain
Session: Modeling and data processing Electromagnetic Techniques 3
Time: 13:30 - 13:50
Online monitoring of material properties is a topic of great interest in many industrial sectors, like the steel industry. Within the Product Uniformity Control project, supported by the European Research Fund for Coal and Steel, industrial and academic partners collaborate to design efficient models able to simulate online systems, which are used within manufacturing processes to non-destructively monitor mechanical properties of steel.
This communication focuses on the IMPOC system, carrying out electromagnetic measurements that correlate very well with microstructure properties. Modelling this system is in itself a difficult task, due to the complexity of material properties involved, which are described at a macroscopic level by a set of magnetic hysteresis curves. Then, simulation studies requiring a large number of calculations bring a better understanding of parameters influence and help correcting possible drifts due to changes in the system environment. Carrying out such studies can quickly become cumbersome, as sensitivity analyses typically require several thousands of model evaluations, for instance. In order to overcome this problem, a strategy based on metamodeling is presented here. It consists in replacing, in a given range of input parameters variation, the numerical solver by real-time approximation obtained by regression over a database of simulation results. Specific tools developed at CEA LIST enable to build such database in an adaptive way, so that the metamodel accuracy is ensured and the number of calculations in the database is minimal. After a presentation of the global strategy for intensive simulation of the IMPOC system, results of simulation studies will be shown and the possible estimation of characteristic parameters from IMPOC signals will be discussed.