Commande prédictive basée sur la simulation. Application à la flottation en colonne

Authors: Bouchard, Jocelyn
Advisor: Desbiens, AndréDel Villar, René
Abstract: Applications of dynamic simulators for model predictive controllers design are rather scarce in the litterature. The complexity of solving the resulting optimization problems may explain this lack of popularity. In fact, nonlinear programming algorithms are not always well suited to efficiently reach the optimum of a fundamentaly-based cost function. The situation is even worse when the equations used in the model are unknown by the control designers (black box models). The simulation-based model predictive controller is an alternative formulation to perform model predictive control (MPC) without making use of any explicit optimization solver, but rather based on an easy-to-compute closed-loop simulation. The resulting scheme generally provides a sub-optimal solution and benefits from many interesting features of conventional MPC without being restricted by the model complexity. Two algorithms are proposed: decentralized and decoupled. The decentralized simulation structure allows a flexible setting of the prediction horizon (Hp) that is not possible in the decoupled case, easier to tune, but where Hp must generally be in the same order of magnitude that the system settling time. A second contribution of this thesis is the development of a framework for the dynamic simulation of a mineral separation process: column flotation. Until now, most of the proposed models or simulators were restricted to the steady-state behavior. When dynamic mass-balance equations were considered, a constant pulp level during the simulation was always assumed. The presented framework aims to simulate water, solids and gas motion and their effect on the pulp level and output flow rates. As it often happens in mineral processing, the column flotation process has not benefited from advanced control techniques. This is where the two previous subjects merge. The proposed simulation framework is used to design a simulation-based model predictive controller for process variables having a strong influence on metallurgical results (grade and recovery). A case study is presented where the pulp level, bias and air hold-up in the pulp zone are kept within an acceptable operating region.
Document Type: Thèse de doctorat
Issue Date: 2007
Open Access Date: 13 April 2018
Permalink: http://hdl.handle.net/20.500.11794/19713
Grantor: Université Laval
Collection:Thèses et mémoires

Files in this item:
SizeFormat 
24892.pdf2.47 MBAdobe PDFView/Open
All documents in CorpusUL are protected by Copyright Act of Canada.