Ph.D.: A Multiple Model Approach for Modeling, Identification and Control of Nonlinear Hybrid Systems
Chemical process plants typically consist of discrete decisions and continuous inputs along with evolution of continuous as well as discrete dynamics. Systems that combine both paradigms namely, continuous state description along with discrete events into a unified framework are known as hybrid dynamical systems. This thesis focuses on the modeling and online control of nonlinear hybrid dynamical systems (NHDS). A novel framework for modeling of nonlinear hybrid systems using multiple partially linearized (MPL) models, is presented. The MPL, which consist of continuous as well as discrete variables, are used to synthesize a model predictive control (MPC) law. The MPC formulation takes on a similar form as that used for control of a continuous variable system. Although implementation of the control law requires solution of an online mixed integer nonlinear program (MINLP), the optimization problem has a fixed structure for any arbitrary NHDS with certain computational advantages. Performance and computational efficiency of the modeling and control scheme is compared with the well known mixed logical dynamical (MLD) model based control. Further, to meet the demand of solving MINLPs online, an efficient optimization algorithm for solving mixed integer nonlinear programs resulting from the MPL model based predictive control formulation for nonlinear hybrid dynamical systems is proposed. The algorithm, based on generalized outer approximation (GOA), uses structural information of the canonical MPL framework as well as analytical expressions for the objective function and constraints. It is shown that the GOA based solution is scalable to larger number of variables and constraints. The proposed modeling, control and optimization algorithms are validated experimentally on three-tank bench mark hybrid system which has been designed and fabricated as part of this thesis work in. Automation laboratory at I.I.T. Bombay. Finally, the above concept extended to data based modeling and a novel identification approach for hybrid dynamical systems which is linear and separable in binary variables is presented. The approach is also validated experimentally on the three-tank experimental setup.
M.Tech.: Phenol Hydroxylation in Liquid Phase
The project was an attempt to develop a process for direct hydroxylation of phenol to hydroquinone. The main aim was to increase the selectivity of hydroquinone and phenol conversion. Zeolites X & Y (catalyst base) were synthesized in the laboratory and various metals (e.g. copper, iron, cobalt, nickel) were impregnated on them. These catalyst were then used in hydroxylation reactions in order improve the selectivity of hydroquinone. It was found that the 0.6% copper impregnated zeolite gave the highest activity compared to other catalysts. Thus, using this catalyst further studies were carried out to study the effect of various parameters such as quantity of catalyst, phenol to oxidant ratio, reaction time etc, on conversion of phenol and yield of hydroquinone.
B.E. : Manufacturing plant design of Phosphoric Acid