Reactor network synthesis with guaranteed robust performance

  • Synthese von Reaktornetzwerken mit robust garantierten Eigenschaften

Zhao, Xiao; Marquardt, Wolfgang (Thesis advisor); Mönnigmann, Martin (Thesis advisor)

Als Manuskript gedruckt. - Düsseldorf : VDI Verlag GmbH (2017)
Book, Dissertation / PhD Thesis

In: Fortschritt-Berichte VDI. Reihe 3, Verfahrenstechnik 952
Page(s)/Article-Nr.: XIV, 171 Seiten : Diagramme

Dissertation, RWTH Aachen University, 2017


Typical continuous process flowsheets include reaction section, separation section and recycles. The reaction section is often the most important part of a chemical process, which may contain several interconnected reactors. The superstructure approach is a widely used model-based process design method for reactor network synthesis. It starts from a reactor network superstructure and uses mathematical models and optimization tools to select the best process design. The superstructure approach results in an optimal process flowsheet with determined connection patterns of reactors, reactor types, design parameters and operating conditions of each reactor. In this work, a systematic model-based approach for reactor network synthesis problems with guaranteed robust dynamic performance will be presented. The work is based on the superstructure approach, but in comparison to the classical methods, not only economic optimality with respect to a static objective function, but also certain specified dynamic properties, i.e. dynamic stability and response speed, are guaranteed simultaneously under parametric uncertainty. Structural alternatives in the flowsheet, i.e., how reactors are interconnected, as well as in the control system, i.e., how controlled and manipulated variables are paired, are subject to design degrees of freedom. Moreover, it is allowed that idle reactors and controllers can appear in the reactor network superstructure, so that a fixed number of non-idle reactors and controllers does not have to be assumed as a priori. The optimal reactor network design in either open- or closed-loop is determined by solving a single optimization problem. The proposed approach allows an integrated treatment of parametric uncertainties, which may either result from model uncertainties, such as reaction kinetic constants or heat transfer coefficients, or from process uncertainties, including slow disturbances in load or the quality of raw materials. A robust eigenvalue constraint to guarantee the robust performance of the designed reactor network is formulated. Efficient formulations of interconnecting reactors and novel complementarity-based constraints for control structure selection are proposed. The method results in a semi-infinite mixed-integer nonlinear optimization problem with complementarity constraints, disjunctions and a robust eigenvalue constraint. A hybrid two-step solution method is proposed to solve the synthesis problem, which integrates candidate solution algorithms of related optimization problems. The proposed solution method is applied to a case study of allyl chloride production with up to ten plug flow and continuous stirred tank reactors.