Optimal analysis and design of flexible and multi-period chemical processes

This thesis addresses the design and synthesis of multiperiod and flexible chemical processes within a mathematical programming framework. Given the model of a process, multiperiod operation requires that the design withstands changes in a set of operating parameters, retaining a feasible operation. The multiperiod design and synthesis problems are formulated as large scale nonlinear programming (NLP), and mixed-integer nonlinear programming (MINLP) problems, respectively, whose solution determines the optimal design, configuration and operation of the process flowsheet. Important aspects of these problems, including computational efficiency, modeling and analysis issues, are investigated.; Firstly, an efficient methodology that includes an NLP and an MINLP algorithm for convex versions of the multiperiod problem is developed. The proposed Outer-Approximation based decomposition is applied to multiperiod multiproduct batch plant problems operating with single product campaigns. Multiperiod models are developed for the design and future capacity expansions of such plants. Numerical results show the method to be advantageous in terms of both robustness and time efficiency.; For the case of nonlinear and non-convex multiperiod design problems an SQP-based, decomposition and projection scheme is developed. Using a quadratic programming subproblem as the main coordination step, the problem in the full space is solved as a stream of independent single period problems. The key property of this method is that the computational effort scales linearly to the number of periods. The ability to solve large scale multiperiod problems is demonstrated on a variety of problems, with superior performance in terms of computational demands, number of function evaluations and solution robustness.; To address uncertainty in the duration of individual periods, a minimax formulation of the objective function is derived. For the analysis of multiperiod problems, a formal definition of bottleneck periods is offered along with a set of theoretical properties that provide an analysis framework that yields a better understanding of multiperiod problems. At the level of computational synthesis and based on these bottleneck properties, a decomposition strategy for general MINLP multiperiod problems is proposed where the MILP part does not scale with the total number of periods.; Finally, a sensitivity based methodology is developed for the flexibility evaluation and design of linear processes. Sensitivity information, derived from the solution of a special form of the imbedded min-max problem, is utilized for the identification and investigation of the supporting active sets. Since the proposed approach automatically generates the supporting active sets it is extended to a methodology for the design of processes with a desirable degree of flexibility. This methodology incorporates analysis and design in a common framework while presenting conceptual and computational advantages that makes it particularly suitable for large scale processes.