Development of a machine learning-based soft sensor for an oil refinery's distillation column

Jimena Ferreira, Martín Pedemonte, Ana Inés Torres

Abstract: Highlights•A novel framework for developing a data-driven soft-sensor is presented.•The framework learns the functional form of the model from real industrial data.•The algorithm combines GP for symbolic regression with statistical methods.•The model outperforms a Kriging based one in validation and degradation scenarios.