| In this research project the question of how process design affects controllability was addressed. In order to compare alternate designs, we developed an analysis tool or index. Our index is independent of the controller and of the disturbance/setpoint change trajectory. Also, it can take into account process constraints, time delays, nonlinearities, and stochastic disturbance models.; The index is defined as the minimum time for the plant to reject/track the worst disturbance/setpoint change with an optimal controller. We define as optimal a controller that at the current time will place the system in such a way that it will take the smallest possible minimum time for the system to recover assuming that any disturbance could have occurred. This controller we call a least worst case controller.; A stochastic model for the disturbances entering the process was used along with a method to discriminate among the disturbances. This model allows for uncertainty in the type, the size, and the time of occurrence of the disturbance. Our model identifier (or discriminator) is based on the Generalized Likelihood Ratio method, and it gives statistical information (i.e. likelihoods and probabilities) which are used to improve our controller. We extended the least worst case controller above in order to include the above probabilities into the formulation.; We have developed an MILP formulation for solving a linear minimum time optimal control problem efficiently and in one pass. Also, we have developed an MILP formulation for solving the least worst case controller for both stochastic and deterministic systems.; In order to show the usefulness of this index and demonstrate the use of the least worst case controller, we assessed the controllability of three secondary reflux vaporization distillation control structures. We successfully applied the probability weighted least worst case controller and showed how well it performs. The results indicated that all three structures perform similarly, with the structure using the heating medium temperature and reflux flowrates being slightly better.; Some preliminary ideas are given on how to use the index in order to retrofit design a plant to improve its quality of control. Finally, suggestions for future work are made. An adaptive multiple model based method is proposed to take into account model uncertainty in our index. We also recommend that the least worst case controller be implemented and tested. |
