Special Joint ChBE/ISR Seminar: Vasilios Manousiouthakis
Monday, September 29, 2014
Room 1146 A.V. Williams Building
Professor Jeffery Klauda
The IDEAS Conceptual Framework and its Application to Attainable Region (AR) Construction for Reactor Networks and General Process Networks
Professor, Department of Chemical and Biomolecular Engineering
Director, Hydrogen Engineering Research Consortium (HERC)
Over the last few decades, computer technology, process modeling, and simulation algorithm advances, have enabled computer based analysis of process networks, thus minimizing pilot plant use for chemical process development. Nevertheless, computer based synthesis of optimal process networks remains limited, and is largely pursued through numerical optimization formulations that are nonlinear or mixed integer nonlinear programs (NLP's or MINLP's) most instances of which cannot be solved globally within realistic timeframes.
The Infinite DimEnsionAl State-space (IDEAS) conceptual framework represents a paradigm shift which establishes that chemical process nonlinearities need not be manifested at the optimal network synthesis level, but rather can be fully accounted for prior to optimization. The resulting mathematical formulations feature feasible regions that are defined by linear constraints, albeit in an infinite dimensional space. Furthermore, for large classes of objective functions, the resulting process network optimization formulations are infinite dimensional linear programs (LP's), whose finite dimensional approximations can be solved to global optimality in a timely manner.
In this talk, the power of the IDEAS framework is illustrated in the quantification of the Attainable Region (AR) concept for reactor networks, and general process networks. The attainable region (AR) for reactor networks was introduced in the 60’s. Since that time, the AR for reactor networks has been the topic of substantial research effort, due to its effectiveness in rigorously assessing, at the early conceptual design stage, the limitations imposed on reactor network performance by the kinetic rates of the underlying reactions, the reactor network feed composition, and the properties of the underlying reactor units. Its quantification aimed to help the design engineer establish rigorous tradeoffs among competing reactor system objectives, independently of any particular reactor network design.
The AR for reactor networks was defined as the complete set of points in concentration space that are product composition vectors of some steady-state reactor networks, using only processes of reaction and mixing/splitting from a given feed point. An overview of “inner” and “outer” methods developed for quantifying reactor network ARs will be given, including the strengths and weaknesses of each method. AR’s will be quantified for isothermal CSTR/PFR networks, non-ideal (dispersion, RTD) reactor networks, reactor networks employing variable-density fluids, and batch reactor networks.
The novel attainable region (AR) concept for general process networks is then presented. This concept is applicable to networks with a finite number of inlets and outlets, and a possibly infinite number of process units, each of which has a finite number of inlets and outlets. The process network AR concept quantifies for the first time the set of all outlet stream composition vectors that can be attained by a process network with inlet/outlet stream flowrate ratio specifications. The process network AR is shown to be a convex set that resides in a concentration space, whose dimension depends on the number of system components and network outlet streams. Quantification of this set is pursued through the IDEAS conceptual framework, and leads to the identification of the vertices of ever more accurate AR approximants. Case studies are presented throughout, to illustrate the underlying concepts and their applications.