Object-oriented Disjunctive Programming For Process Synthesis

Solving process synthesis problems by mathematical programming methods has attracted great attention during the last four decades in both research area and application domains. As a more advanced modeling alternative for MINLP in process synthesis, the generalized disjunctive programming (GDP) was proposed in the 1990s. In this paper a novel object-oriented disjunctive programming (ODP) model with a solver is proposed. Applications have been conducted to various process synthesis areas. The work in this paper includes the following contents:(1) An improved variant of the traditional GDP, otherwise known as the object-oriented disjunctive programming, is proposed. By introducing the object-oriented programming idea, it can eliminate all the redundant variables and terms in the generated sub-NLPs during the solving iteration.(2) A hybrid method by combining the heuristic algorithm and gradient-based optimizer is proposed to solve ODP as well as GDP models. As a two-layer method, a heuristic algorithm performs master iterations in the outer-loop searching for the global optimal structure, and a gradient-based NLP solver is applied in the inner-loop solving the sub-NLP when the structure is determined by the outer-loop.(3) The GDP and ODP models of general heat exchanger network synthesis (HENS) are proposed. The hybrid method is successfully applied to medium and large scale HENS problems. Numerical results demonstrate the effectiveness of the ODP modeling and hybrid method. The advantages of ODP over GDP and MINLP are also demonstrated via case studies.(4) Typical water network synthesis (WNS) problems have been modeled in the ODP formulation and solved by the hrbrid method. The advantages of ODP over GDP in this area are also revealed.