Experiences in applying data-driven modelling technology to steelmaking processes

Abstract

Experience has shown that data-driven modelling methods are useful for improving steelmaking processes. In particular, principal components analysis and partial least squares are well-suited for industrial implementation because they address practical issues such as colinearity and missing data. In the course of applying these multivariate methods on-line, a need for a flexible computer infrastructure to better support data handling and model implementation was identified and met with an internally developed software calculation platform. Multivariate methods have been found useful for monitoring and for prediction and can also be applied as a foundation for other methods such as optimization. © 2008 Canadian Society for Chemical Engineering.