Multiperiod chemical batch plant design for a single product or mixed product campaign mode of operation

Design models for multiproduct batch plants define the optimal size and number of batch equipment needed to produce predetermined product quantities over a strategic planning horizon, and such for every stage of the production process. However, although these design models incorporate realistic equipment costs and process information, a realistic spread of product deliveries over the (long) planning horizon, is mostly not considered. In single period design models, product deliveries are not modelled: it is assumed that finished products can be stored (for free) until delivery, or are delivered directly after production. Multiperiod plant design models assume in general that product quantities are delivered at the end of every repetitive (short) production period, but the influence of a variable delivery scheme has barely been investigated. The aim of this paper is to study the design of a batch plant for a periodic delivery scheme with variable product quantities, and compare the results for two modes of operation: single product and mixed product campaigns, both in a multiperiod context. Additionally, we will evaluate the impact of (limited) storage capacity for finished products and variable product mixes over the periods. All appropriate batch plant design models are solved with mixed integer techniques. The objective is to minimise capital and startup costs, while inventory holding costs are calculated ex post. From the exploratory study performed, it appears that the cost of periodic deliveries with variable product quantities, seems not that high, especially when end-of-period inventory is allowed. Concerning the mode of operation, the mixed product campaign mode turns out to be very complex to calculate, especially in a multiperiod context, without distinct cost advantages for the plant. Therefore we conclude that, if feasible for the production process, producing in multiperiod single product campaigns is often the cheapest way to operate a batch plant that allows a variable delivery scheme.