Scalable model for industrial coffee roasting chamber

Abstract

The temperature profile of the coffee beans during the roasting phase determines the colour, aroma and flavour of the coffee. In order to reproduce these desired characteristics, the control of the coffee beans temperature has a key role in the roasting process. A proper model of the plant is required to design an intelligent control. Recently, several physical models that share the main physical equations have been proposed, but they have physical parameters specific of each process. In such scenario, each plant requires an ad hoc identification of the model parameters. This work proposes a model of the roasting chamber that can be used on plants of different sizes by scaling only geometrical parameters directly measurable on the roasting plant. The proposed model was identified on a 120 kg plant and then applied to a 360 kg one. The obtained results show in both cases similar accuracy (FIT = 75.49%, MPE=4.66%).