| Batch optimization has been an area of active research in recent years. Numerous methods have been developed, and while many have met with success, serious shortcomings remain. No methods exist to generate continuous optimal profiles for batch processes in a true on-line fashion without prior knowledge of the profile structure.; In this work, a methodology for designing and implementing a real time optimization for nonlinear batch processes is presented. The system input and state trajectories are determined on-line to minimize a cost function. An interior point method with penalty function is used to incorporate constraints into a modified cost functional. A Lyapunov based adaptive gradient approach is used to compute the trajectory parameters. This technique is first proposed for optimizing differentially flat systems in a cascade implementation. This theory is then expanded to a general nonlinear control system. Smooth trajectories are generated with feasible computing time compared to many optimization methods that have been proposed in literature.; Cascade optimization optimizes the profiles over the whole batch, and uses measurement feedback in the form of a controller. Controller design may be difficult, so it is necessary to incorporate the measurement feedback into the optimization itself. A true on-line optimization is developed that optimizes the profiles over the remaining portions of the batch. This gives a more practical application that can be applied to a general nonlinear system, with no control design necessary. Local optimization is obtained with no knowledge of the optimal profile structure. |
