MILP Modelling of Production-related Disturbances

Abstract: Rising cost awareness within the process industry has emphasized the need for better scheduling and modelling of manufacturing sites. In collaboration with the Process Industrial Centre at Lund University and Perstorp AB, this study aims to model utility disturbances in manufacturing sites. Utilities are services or equipment necessary for the production in one or several of the areas that a site consists of. The goal of the modelling is to minimise economic losses related to utility disturbances. Manufacturing sites can be mapped as area networks with interdependent areas, buffer tanks and utilities that have defined capabilities. In this study a general method for modelling area networks as mixed integer linear programs (MILP) have been presented. If the utility disturbances length and scope are unknown it is possible to use stochastic programming in the model, which would enable minimisation of losses by optimising over a number of scenarios, or guesses, of how a disturbance will unfold. To demonstrate how a site can be represented by a MILP-model, an area network corresponding to a part of the Perstorp AB Stenungsund site has been modelled. The main result,from optimising the modelled site are production rates for the areas, which would minimise the economic losses from a specific disturbance.