Nowadays, industrial processes must be efficient, particularly at the production level and/or energy consumption. This research work aims at improving the process efficiency by analysing the influences of disturbances on their behaviour, from the conception phase to the synthesis of controller/observer, in an integrated approach. The disturbance rejection problem is first introduced as well as different control laws allowing attenuate/reject these disturbances. The Torsion-Bar (T-B) system is presented. It will be used as case study all along this research work. A control law based on the concept of derivative state variable is presented and validated while applied as disturbance rejection on the T-B system. In order to reject the disturbance, different physical variables must be estimated, such as state variables, derivative state variables as disturbance variables. An unknown input observer based on the bond graph representation is recalled and extended in the multivariable case. It is the first theoretical contribution of this work. As for the synthesis of control laws, an integrated approach is developed. We thus compare the efficiency of different so-called « modern control laws » for the disturbance rejection problems by simulation with the T-B system example. We analyse the efficiency of our approach. One extension to the Input-Output decoupling problem allows us to extend the disturbance rejection problem to other control law type in an integrated approach. At least, these techniques are applied on the real T-B system and compared. We validate our approach. Since this work aims at analysing and developing efficient control laws for industrial processes, a simplified model of a hydroelectric plant is developed, in order to apply our results. A simplified bond graph model is validated with simulations. It will be used as a basis for the control of a plant. We are working with researchers of the fluid mechanic lab for developing a first plant.
Directeur de thèse : M. SUEUR Christophe, Professeur, Ecole Centrale de Lille Rapporteurs : M. HAMELIN Frédéric, Professeur, Université de Lorraine M. MANAMANNI Noureddine, Professeur, Université de Reims Champagne-Ardenne Examinateurs : Mme. FANTONI Isabelle, Directeur de Recherche, Ecole Centrale de Nantes M. MERZOUKI Rochdi, Professeur, Polytech Lille M. DUPONT Patrick, Professeur, Ecole Centrale de Lille Invitée : Mme. DAUPHIN-TANGUY Geneviève, Professeur Emérite, Ecole Centrale de Lille
Thèse de l'équipe MOCIS soutenue le 02/12/2019