The aim of this thesis is to improve the overall efficiency of the multiple hearth furnace (MHF) in kaolin calcination by developing control strategies which incorporate machine learning based soft sensors to estimate mineralogy related constraints in the control strategy. The objective of the control strategy is to maximize the capacity of the furnace and minimize energy consumption while maintaining the product quality of the calcined kaolin. First, the description of the process of interest is given, highlighting the control strategy currently implemented at the calciner studied in this work. Next, the state of the art on control of calcination furnaces is presented and discussed. Then, the description of the mechanistic model of the MHF, which plays a key role in the testing environment, is provided and an analysis of the MHF dynamic behavior based on the industrial and simulated data is presented. The design of the mineralogy-driven control strategy for the multiple hearth furnace and its implementation in the simulation environment are also outlined. The analysis of the results is then presented. Furthermore, the extensive sampling campaign for testing the soft sensors and the control strategy logic of the industrial MHF is reported, and the results are analyzed and discussed. Finally, an introduction to Model Predictive Control (MPC) is presented, the design of the Linear MPC framework for the MHF in kaolin calcination is described and discussed, and future research is outlined.
|Tila||Julkaistu - 2020|