| This thesis focuses on two main areas of research, namely: dynamic optimization of non-steady state processes and the use of nonlinear control, in a differential geometric setting, in controlling reactor temperature and production rate in gas-phase polyethylene reactors.; The use of iterative dynamic programming (IDP) in obtaining reliable optimal setpoint trajectories for non-steady state operations is investigated. A novel approach for selecting penalty functions in handling continuous and final state constrained problems is proposed. For systems with time-delay in the state, a quadratic approximation of the delayed state is proposed. In addition, a first order Taylor series approximation has been proposed to handle systems with deadtime, resulting in a piecewise linear optimal control profile, instead of a piecewise constant profile of same.; A new simultaneous iterative solution, which is computationally attractive, has been proposed for handling time-optimal control problems arising in chemical engineering. It is shown, in this thesis, that many of the gradient-based methods used for obtaining singular sub-arcs may not be well-suited and that direct search methods may be better in obtaining optimal trajectories for such problems.; Nonlinear error trajectory controllers (ETC), based on a differential geometry framework, are used to show the potential benefits of employing nonlinear control strategies for systems which are severely nonlinear. A gas-phase polyethylene reactor (GPPER) system, which is prone to open-loop temperature oscillations and exhibits steady state multiplicity, was studied extensively in this thesis. A tuning approach for minimum phase systems of arbitrary relative degree has been proposed. In addition, the impact of the heat exchanger model has been addressed.; The effect of controller saturation, due to physical constraints, has been addressed in this work. Sufficient conditions for the existence of estraneous equilibria have been established. Finally, the interaction between production rate optimization scheme and reactor temperature control scheme is addressed. It is shown that a multivariable controller has no advantage over multiloop controllers, suggestive of the fact that process nonlinearity is more important to address than the interaction. |
