Optimization based methodology for refrigeration system synthesis and molecular design.

摘要

This thesis describes systematic methods for (i) synthesizing refrigeration systems involving pure and mixed refrigerants and (ii) designing molecules with target properties correlated with topological indices. Since the underlying question in both the problems is to efficiently search through a large number of alternatives to find the best solution, this work models each problem within an optimization framework.; The first part addresses the optimal synthesis of the refrigeration configuration and the selection of the best refrigerant(s) (pure) from a list for satisfying a set of process cooling duties at different temperatures. While existing practice is to perform refrigerant selection largely in isolation of refrigeration cycle synthesis it is shown that there is a clear advantage in performing both tasks at the same time. To this end, an optimization-based approach is described that simultaneously selects refrigerants and synthesizes refrigeration structures by minimizing a weighted sum of investment and operating costs.; Next, the synthesis of refrigeration systems with refrigerant mixtures as working fluids is addressed. The employed refrigeration topology encompasses features from industrial (Liquefaction of Natural Gas) LNG systems. This configuration consists of a combination of horizontal and vertical cascades generalizing the vertical cascade of pure refrigerant systems. The key features of mixtures exploited here are their ability to evaporate/condense over a temperature range and their potential to generate streams with different compositions through partial condensation. The synthesis problem is formulated and solved as a nonlinear program.; The final part addresses the problem of designing product molecules with fine—tuned or optimized property values using structure—property relations with topological indices as structural descriptors. The topological indices considered are Randić's molecular connectivity indices, Kier's shape indices and the Wiener Index. The adjacency matrix representation which provides a complete description of the connectivity of a molecule is utilized. The nonlinear expressions for the topological indices are systematically transformed into equivalent linear relations. This enables the formulation of the molecular design problem as a Mixed Integer Linear Program (MILP).