Radoslav Paulen will present his defence of doctoral minithesis entitled Global Dynamic Optimization of Processes.
Abstract: This work concerns problem of global and dynamic optimization of continuous and hybrid processes. Hybrid processes keep both discrete and continuous character and they provide better dynamic information of phenomena appearing in real world. So hybrid process models are very useful for modeling of processes with not simple dynamics (e.q. processes which cannot be discribed by simple ordinary or partial differential equations). Dynamic optimization (DO) represents set of mathematical methods utilized to compute optimal (open–loop) control for dynamic process. From various DO methods we choose control vector parametrization method (CVP) which discretize a continuous control to sum of basis functions (like piece–wise constant functions, linear functions, or polynomial functions) a then forms non–linear programming (NLP) problem. This NLP problem can be handled efficiently by any of numerous NLP solvers available if accurate gradient information is provided. Gradient computation is performed employing method of adjoint variables. Since DO problem may possess multiple (local) optima, methods of global optimization must be applied to determine absolute (global) optimum. In this work we discuss usage of method of global and dynamic optimization for hybrid process of emulsion copolymerization of styrene and α–methyl styrene.