Design of pipeless chemical batch plants with queueing networks

Abstract

The problem of designing pipeless batch plants is tackled using a multiclass re-entrant queueing network model. The overall design procedure is divided into three hierarchical phases: the preliminary phase, neighbor search algorithm (NSA), and greedy mean value analysis (GMVA). In the preliminary phase, the types of products and stations, as well as the capacities of equipment items, are determined. GMVA, which is a generalization of conventional MVA, finds the optimal job constitution for the system within a short computer processing unit (CPU) time, after the station configuration is fixed, whereas the station configuration is successively improved by NSA. The moderate time complexity of the proposed method extends the range of applications to practical-sized problems. An alternative design approach that is based on the decomposition approximation and formulated as a mixed integer nonlinear programming (MINLP) problem is also developed. The effectiveness of the method is illustrated through the examples of mediumsized pipeless plant design problems. Simulation studies using a commercial simulation package are described and observed to support the validity of the analytic method.