Iranian Journal of Chemical Engineering (IJChE)

Abstract Common methods to design heat exchanger networks (HENs) by pinch technology usually need an evolutionary step to reduce the number of heat transfer units. This step is called loop breaking and is based on the removal of exchangers that impose minimum increase on utility consumption. Loops identification and breaking is a tedious task and becomes more complicated in large networks. This paper presents a rapid nonlinear programming (NLP) formulation for the evolution of HENs in which loop identification is not required. The objective of the NLP is the minimization of HENs annual cost, which is not considered in current methods. In this method a search is done to find the best units elimination of which improves HENs annual cost. The search continues until the minimum number of units (MNU) is achieved and the exchangers that must be removed from the network are specified. The method was applied to some networks reported in the literature and better results were obtained. Also, the convergence of the presented method is very fast and it can be applied to different networks designed by pinch technology.

Abstract Synthesis of heat exchanger networks (HENs) is inherently a mixed integer and nonlinear programming (MINLP) problem. Solving such problems leads to difficulties in the optimization of continuous and binary variables. This paper presents a new efficient and robust method in which structural parameters are optimized by genetic algorithm (G.A.) and continuous variables are handled due to a modified objective function for maximum energy recovery (MER). Node representation is used for addressing the exchangers and networks are considered as a sequence of genes. Each gene consists of nodes for generating different structures within a network. Results show that this method may find new or near optimal solutions with a less than 2% increase in Hen annual costs.