Multivariate process monitoring of EAFs

The application of a multivariate statistical process control (MSPC) to an EAF and the benefits that can be delivered was discussed. Several statistical methods for multivariate prediction were tested such as multiple linear regression (MLR), principal component regression (PCR) and partial least squares (PLS). The results show that PLS was the most suitable of the tested methods and the prediction accuracy for tramp elements and alloying elements were satisfactory for online predictions and condition monitoring of scrap properties. Monitoring of short and long term variations in scrap quality was possible by analysis of the prediction errors and regression coefficients.