A Comparative Study of Deterministic and Stochastic Optimization Methods for Integrated Design of Processes

This paper focuses on the application of stochastic (genetic algorithms, simulated annealing) and deterministic (sequential quadratic programming) optimization methods for the integrated design of processes considering dynamical non-linear models. Moreover, a hybrid methodology that combines both types of methods is proposed, showing an improvement on performance. Controllability indexes such as disturbance sensitivity gains, the H infinity norm, and the ISE were considered to obtain optimum disturbance rejection. In order to illustrate and validate our proposal, an activated sludge process with PI schemes is taken. The problem is stated as a multiobjective non-linear optimization problem with non-linear constraints. The application of the mentioned strategies is discussed. The results are encouraging for future application of these techniques to solve synthesis MINLP problems