A comparison of modelling approaches for closed-loop decision making over discrete domains in manufacturing systems

The manufacturing industry is transforming towards smart, flexible, and energy-efficient systems. In this regard, control strategies based on optimisation have been proposed to improve the energy efficiency of manufacturing systems. Usually, the optimisation problem behind the controller design involves decision variables constrained to binary and discrete domains resulting in Mixed-Integer Linear Programming (MILP) problems and require a high computational burden to find their optimal solution. In this paper, three different approaches are proposed for the closed-loop decision making over discrete domains in manufacturing systems, including the constraints and limitations to their implementation. Thus, the number and nature of required variables as well as the additional constraints needed to represent the discrete domains of the decision variables are established. These approaches are tested and compared to solve the optimisation problem behind a predictive-like controller designed to minimise the energy consumption of a manufacturing process line.