摘要：
The chemical industry in the United States can be cited for numerous incidents of overdesign, incidents that increasingly threaten our ability to produce chemicals and chemically related products competitively with foreign manufacturers. Usually, overdesigns compensate for model uncertainties, allow for increases in capacity, and provide margins of safe operation. Less obvious are overdesigns that circumvent operation near or within complex operating regimes. However, with the development of more reliable model predictive controllers (MPCs), an important opportunity and challenge faces process designers: how to utilize nonlinear analysis to obtain more economical designs that are flexible and controllable in regimes characterized by greater sensitivities to modeling errors (process/model mismatch) and changes in set-points, and in which good control is more difficult to achieve. This thesis presents a prototype computing system (PRODOC - PROcess Design, Operations, and Control) for the design and analysis of such highly nonlinear processes. Within this coordinated framework are several contributions, including a novel model predictive control algorithm that provides excellent servo- and regulatory control for nonlinear processes and a new optimization algorithm that transforms constrained, nonlinear programs (NLPs) to systems of nonlinear equations (NLEs) that are solved using robust homotopy-continuation algorithms. In the strategy presented, as the design optimization proceeds, the performance of the nonlinear MPC is evaluated in response to typical disturbances and the design objective function is penalized for poor controllability. MPCs have been shown to provide the best response to unmeasured disturbances, permit the inclusion of constraints, and be fairly insensitive to their tuning parameters, thus possibly allowing for an evaluation of the inherent controllability of the process. This strategy has been applied to a bioreactor process that exhibits complex c
欢迎来到三峡大学图书馆！
