Using neural networks to account for cross-zone interactions in metal reheating furnaces

Abstract

The study focuses on the development of a functional model describing cross-zone interactions within a reheating furnace used prior to rolling. Various methods for the identification of cross-zone interaction model, ranging from regression analysis to neural networks were described. A neural network model that reflects the state of the furnace zones while accounting for their mutual interactions was obtained. The overall regression coefficient (across all zones) was approximately 0.82, with individual zone coefficients ranging from 0.65 to 0.85. Cross-zone interactions were modeled, and the results were validated by expert evaluation.