(389b) Constrained Optimal Particle Size Distribution Control in a Pharmaceuticals Batch Dryer Using a Hybrid Genetic Algorithm-Pattern Search Optimization Strategy

This contribution describes the dynamic optimization of a pharmaceuticals batch dryer. The optimal control problem is based on the control vector discretization or single shooting method, thus a non-linear programming (NLP) problem is obtained. The target of the open loop optimization problem is the batch time minimization and the control variables are the On/Off intervals of the mixer, and the mixer rotation speed. In order to implement a relevant scenario from practical point of view the following constraints are applied: the on and off intervals have same duration during the entire batch, the stirrer speed is constant, and the final moisture specification must be met. Finally, an upper limit on the fines fraction of the particle size distribution is set. Due to these constraints the optimization problem presents several local minima and low-gradient regions, thus a gradient based method fails to converge. Therefore, a hybrid, genetic algorithm-pattern search (simplex) method is used in an alternating (global-local-global-local) search implementation.