Scheduling of multipurpose batch chemical plants

Michael Charles Wellons, Purdue University

Abstract

Production in a multipurpose batch chemical plant can be achieved in a series of campaigns, where each campaign consists of one or more independently operated single-product production lines. As a consequence, the problem of scheduling multipurpose batch plants can be decomposed into three subproblems: a production planning problem in which the available production time is allocated to a set of alternative campaigns, a campaign formation problem in which good configurations of the plant equipment items are generated, and a production line scheduling problem in which the optimal schedule is identified for a particular equipment configuration. A mixed-integer nonlinear programming (MINLP) formulation for the production line scheduling problem is developed which incorporates equipment group sequencing and variable batch sizes. A combinatorial analysis shows that many of the feasible solutions result in equivalent schedules. An enumeration procedure is developed to identify the operationally and rotationally distinct path sequences for the production line. Procedures are developed to incorporate this information directly into the MINLP formulation so that an O(NP) fewer alternatives need to be considered. A MINLP formulation for the campaign formation problem is developed in which the production line scheduling model is augmented with constraints on the equipment assignment. A decomposition solution strategy is developed which alternately solves an equipment group master problem to identify a good equipment group profile and a campaign formation subproblem which identifies the best campaign for a particular equipment group profile. For single-product campaigns, the production line with highest processing rate is identified. For multiple product campaigns, a small set of dominant campaigns is generated based on a linear dominance property. The dominant campaigns comprise the noninferior extreme points of the associated multiobjective campaign formation problem. The noninferior set estimation method is incorporated into the decomposition strategy to identify these campaigns. Finally, a multi-period MILP production planning formulation for multipurpose batch plants which accounts for lost production time due to changeovers and startup times is developed. Application of the entire scheduling methodology to the Mauderli and Rippin multipurpose batch plant yields a 5% improvement in profit as compared with previous scheduling methodologies.

Degree

Ph.D.

Advisors

Reklaitis, Purdue University.

Subject Area

Chemical engineering

COinS