9706950 Biegler This proposal deals with the development and application of interior-point methods (IPMs) to nonlinear programming (NLP) problems. Particular attention will be devoted to large-scale and structured quadratic programming problems that are encountered as subproblems in the solution of NLPs coming from chemical process engineering. The proposed research will focus on exploiting the structure of various classes of process engineering problems through efficient decomposition methods in large-scale linear algebra. In addition, principal investigators will develop and refine an interior-point strategy that has close parallels with successive quadratic programming (SQP) algorithms and address some open questions related to the efficient use of warm-starts in IPMs. Finally, these algorithms will be implemented to demonstrate their performance on large-scale problems in process engineering. Increased international competition along with environmental constraints and resource limitations require much more sophisticated design and manufacturing strategies for chemical process industries. Over the past decade these needs have started to be addressed by efficient optimization strategies. Nevertheless, the size and complexity of these problems (with sizes approaching a million variables) impose a heavy burden on current optimization algorithms. The proposed research will allow the solution of much larger optimization problems faced by this industry. Through the development and application of interior point methods for nonlinear programming, we will be able to develop tailored solution strategies for the optimization of large-scale steady state and dynamic process models. This will lead to consideration of much larger and more difficult process engineering models, as well as the integration of multiple processes. The result will lead to chemical processes that are environmentally benign, very efficient in the conversion of raw materials to products and hi ghly competitive in today's marketplace.
