Nonsmooth Analysis In Process Modeling, Design And Optimization

Abstract

Nonsmooth models provide a powerful, compact and intuitive modeling paradigm that can represent a growing spectrum of physical phenomena over broad ranges of conditions in terms of the original process variables. This paper advocates for and illustrates the potential of nonsmooth steady-state models in a variety of process engineering applications. Although less general than, for example, mixed-integer formulations, the nonsmooth approach can capture many of the traditional chemical engineering applications of mixed-integer models. Advantageously, the nonsmooth approach supports the computation of useful local sensitivity information, and the formulation of currently challenging problems as simulations (equation-solving) or local optimizations with compact models, so it can scale well to large and complex applications. This potential is illustrated by detailed examinations of liquefied natural gas process design, and process integration problems targeting multiple resources coupled with detailed process models.