Optimal Scheduling of Multistage Polymer Plants

This article introduces a new mixed integer linear programming model of the optimal scheduling of multistage polymer plants. Most commodity polymer plants for products such as HDPE, PP, LDPE, LLDPE, PVC, PS, K-RESIN and CB are composed of reaction, extrusion and packaging stages. The operation of the reaction stage has the characteristics of both a batch process in that one process produces multiple products in sequence as well as a continuous process in that the process runs continuously during product transition and the lot size for each product is variable. The operating mode of the extrusion and the packaging stages are semi-continuous. The scheduler should reduce the number of product transition in order to reduce the sequence dependent product changeover cost in the reaction stage. However, reduction of the number of product changeovers in the reaction stage can cause the overstocking of same products and shortage of the other products in the final packaging stage. These consequences increase costs such as inventory holding cost and back-logging cost of packaged products due to late delivery. The development of mixed integer linear programming models for optimizing a single production stage has been reported in the literature (see, for example, Karimi and Mcdonald, 1997). In this paper, we exhibit the model to include the extrusion and packaging stages into consideration under mild assumptions. The resulting large-scale model and solution strategy is tested with actual operation data collected from 5 plants.