Bayesian Mechanical Property Estimation for Stable Rolling Mill Processes

This study presents a Bayesian approach using Gaussian processes to estimate mechanical properties in stable rolling mill processes. The meaning of process stability is stable plant setup for a given steel grade, a common characteristic of industrial productions. By using a reduced number of plant and metals characteristics, the model shows its robustness across the various steel grades. The huge amount of available data enhances model performance and generalization. Moreover, a method to compute an optimal chemical composition was designed in order to avoid extrapolation on available data.

Authors:

Jacopo Piccirillo | Danieli Automation S.p.A.
 

Sara Marzio | Acciaierie Bertoli Safau S.p.A.
 

Session Chairs:

Tucker Keuter | Charter Steel - Saukville, Wisconsin
 

Shaojie Chen | EVRAZ Regina